oEPA
Office of Water	EPA-822-R-07-010
Environmental Protection	4304T	December 2016
Agency
Public	Review
Field-Based Methods for Developing Aquatic
Life Criteria for Specific Conductivity
U.S. Environmental Protection Agency
Office of Water, Washington, DC

-------
DRAFT
DO NOT CITE OR QUOTE
December 2016
Public Review Draft
PUBLIC REVIEW DRAFT
Field-Based Methods for Developing Aquatic Life Criteria
for Specific Conductivity
U.S. Environmental Protection Agency
Office of Water
Office of Science and Technology
Washington, DC
11

-------
CONTENTS
LIST OF TABLES	viii
LIST OF FIGURES	x
LIST OF ABBREVIATIONS AND ACRONYMS	xiv
NOTICES	xv
FOREWORD	xvi
AUTHORS, CONTRIBUTORS, AND REVIEWERS	xvii
EXECUTIVE SUMMARY	xix
GLOSSARY	xxii
1.	INTRODUCTION AND BACKGROUND	1-1
2.	PROBLEM FORMULATION	2-1
2.1.	PROBLEM IDENTIFICATION	2-1
2.2.	STRESSOR 01 CONCERN—SALTS	2-2
2.2.1.	Sources of Ions	2-3
2.2.2.	Conceptual Model	2-8
2.2.3.	Environmental Transport and Fate of Ions in the Aquatic Environment	2-9
2.3.	MEASURE OF EXPOSURE	2-11
2.4.	NATURE OF THE EFFECT	2-13
2.5.	MECHANISMS AND MODES OF ACTION	2-16
2.5.1.	Physiological Mechanisms	2-17
2.5.2.	Mortality, Growth, and Reproduction	2-19
2.5.3.	Emigration	2-19
2.5.4.	Failure to Recruit	2-19
2.5.5.	Community Interactions	2-20
2.6.	ASSESSMENT ENDPOINTS AND MEASURES OF EFFECT	2-20
2.6.1.	Assessment Endpoints	2-20
2.6.2.	Measures of Effect	2-22
2.7.	SELECTION OF A FIELD-BASED METHOD	2-25
3.	ANALYSIS PLAN: FIELD-BASED METHODS TO DEVELOP SPECIFIC
CONDUCTIVITY CRITERIA	3-1
3.1. DERIVING A CRITERION CONTINUOUS CONCENTRATION (CCC)	3-1
3.1.1.	Establishing the Data Set	3-4
3.1.1.1.	Information Sources	3-4
3.1.1.2.	Selection and Adequacy of Data Sets	3-5
3.1.1.3.	Quality Assurance/Quality Control (QA/QC)	3-10
3.1.2.	Calculating Genus Extirpation Concentrations (XC95)	3-11
3.1.2.1. Assigning Qualifying Designation to Extirpation
Concentration (XC95) Values	3-14
3.1.3.	Calculating the Community-Level Effect Estimate Hazardous
Concentration (HC05)	3-15
3.1.3.1. Validating the Effect Estimate Hazardous Concentration
(HC05) by Bootstrapping	3-16
111

-------
CONTENTS (continued)
3.1.4. Assessing Seasonality, Life History, and Sampling Methods	3-18
3 .2. DERIVING A CRITERION MAXIMUM EXPOSURE CONCENTRATION
(CMEC)	3-20
3.3.	ESTIMATION OF CRITERIA DURATION	3-22
3.3.1. Summary of Recommended Duration for Criterion Continuous
Concentration (CCC) and Criterion Maximum Exposure Concentration
(CMEC)	3-25
3.4.	ESTIMATION OF CRITERIA FREQUENCY	3-25
3.4.1.	Recovery Rates in Literature Reviews	3-27
3.4.2.	Life History Considerations	3-29
3.4.3.	Summary for Field-Based Frequency for Criterion Continuous
Concentration (CCC) and Criterion Maximum Exposure Concentration
(CMEC)	3-30
3.5.	ASSESSING CAUSATION	3-30
3.6.	ASSESSING WATERBODY APPLICABILITY	3-32
3.6.1.	Stream pH	3-32
3.6.2.	Waterbody Type	3-32
3.7.	METHODS FOR APPLICATIONS TO NEW AREAS	3-35
3.7.1.	Application within an Ecoregion—Background Matching	3-36
3.7.1.1.	Obtaining a Data Set	3-40
3.7.1.2.	Estimating Background Specific Conductivity (SC)	3-40
3.7.1.3.	Background-Matching Approach	3-41
3.7.1.4.	Options When Background in New Area is Different than in
Original Area	3-42
3.7.1.5.	Summary of B ackground Matching Method	3 -42
3.7.2.	Developing a Criterion Using Background-to-Criterion Regression
Method	3-43
3.7.2.1.	Using Background to Calculate a Hazardous Concentration
of the 5th centile (HCos) of a Taxonomic Extirpation
Concentration Distribution (XCD)	3-46
3.7.2.2.	Formula for Calculating the Hazardous Concentration of the
5th centile of a Taxonomic Extirpation Concentration
Distribution (HCos) from the Background-to-Criterion Model.... 3-48
3.7.2.3.	Formula for Calculating the Lower and Upper 10%
Prediction Limits	3-48
3.7.2.4.	Criterion Continuous Concentration (CCC) with <200 Paired
Biological Data	3-50
3.7.2.5.	Criterion Continuous Concentration (CCC) with 200 to 500
Paired Biological Data	3-50
3.7.2.6.	Calculation of the Criterion Maximum Exposure
Concentration (CMEC)	3-50
3.7.2.7.	Summary	3-50
iv

-------
CONTENTS (continued)
4. CASE STUDY I: EXAMPLE USING EXTIRPATION CONCENTRATION
DISTRIBUTION (XCD) METHOD IN A LOW BACKGROUND SPECIFIC
CONDUCTIVITY ECOREGION	4-1
4.1.	DATA SET CHARACTERISTICS	4-1
4.1.1.	Background Specific Conductivity	4-6
4.1.2.	Ionic Composition	4-9
4.1.3.	Seasonal Specific Conductivity Regime	4-11
4.2.	RESULTS	4-12
4.2.1.	Extirpation Concentration (XC95) and Hazardous Concentration (HC05)
Values (Example Criterion Continuous Concentration [CCC])	4-12
4.2.2.	Example Criterion Maximum Exposure Concentration	4-15
4.3.	GEOGRAPHIC APPLICABILITY	4-17
4.3.1.	Data Sources	4-18
4.3.2.	Geographic Applicability Results	4-20
4.4.	SUMMARY OF EXAMPLE CRITERIA FOR ECOREGION 69	4-22
4.5.	EXAMPLE CRITERION CHARACTERIZATION	4-23
4.5.1.	Factors Potentially Affecting the Extirpation Concentration
Distribution (XCD) Model	4-23
4.5.1.1. Sensitivity Analyses	4-23
4.5.2.	Validation of the Extirpation Concentration Distribution (XCD) Model.... 4-25
4.5.3.	Duration and Frequency	4-26
4.6.	PROTECTION OF FEDERALLY-LISTED SPECIES	4-27
5. CASE STUDY II: EXAMPLE USING THE EXTIRPATION CONCENTRATION
DISTRIBUTION (XCD) METHOD IN A MODERATE BACKGROUND
SPECIFIC CONDUCTIVITY ECOREGION	5-1
5.1.	DATA SET CHARACTERISTICS	5-1
5.1.1.	Background Specific Conductivity	5-6
5.1.2.	Ionic Composition	5-9
5.1.3.	Seasonal Specific Conductivity Regime	5-10
5.2.	RESULTS	5-12
5.2.1.	Extirpation Concentration (XC95) and Hazardous Concentration (HC05)
Values (Example Criterion Continuous Concentration)	5-12
5.2.2.	Example Criterion Maximum Exposure Concentration	5-15
5.3.	GEOGRAPHIC APPLICABILITY	5-17
5.3.1.	Data Sources	5-18
5.3.2.	Geographic Applicability Results	5-20
5.4.	SUMMARY OF EXAMPLE CRITERIA FOR ECOREGION 70	5-22
5.5.	EXAMPLE CRITERION CHARACTERIZATION	5-22
5.5.1.	Factors Potentially Affecting the Extirpation Concentration
Distribution (XCD) Model	5-22
5.5.1.1. Sensitivity Analyses	5-22
5.5.2.	Validation of the Model	5-24
v

-------
CONTENTS (continued)
5.5.3. Duration and Frequency	5-25
5.6. PROTECTION OF FEDERALLY-LISTED SPECIES	5-26
6.	CASE STUDY III: EXAMPLE USING THE BACKGROUND TO CRITERION
(B-C) REGRESSION METHOD	6-1
6.1.	DATA SET CHARACTERISTICS	6-1
6.1.1.	EPA-Survey Data Set	6-3
6.1.2.	U.S. Geological Survey (USGS) Data Set	6-5
6.1.3.	Modeled Mean Baseflow Background Specific Conductivity (SC)	6-8
6.1.4.	Characterization of Ionic Matrices	6-8
6.1.4.1.	EPA-Survey Data Set Ionic Characteristics	6-8
6.1.4.2.	U.S. Geological Survey (USGS) Data Set Ionic
Characteristics	6-9
6.1.5.	Comparison of Background Specific Conductivity (SC) Estimates	6-9
6.1.6.	Calculation of Ecoregion 43 Criterion Continuous Concentration
(CCC) from Background	6-11
6.1.7.	Formula for Calculating the Lower 50% Prediction Limit	6-13
6.1.8.	Example Criterion Maximum Exposure Concentration	6-15
6.2.	EXAMPLE CRITERION CHARACTERIZATION FOR ECOREGION 43
BASED ON A BACKGROUND-TO-CRITERION MODEL	6-17
6.3.	PROTECTION OF FEDERALLY-LISTED SPECIES AND OTHER
HIGHLY VALUED TAX A	6-18
7.	CASE STUDY IV: EXAMPLE USING THE BACKGROUND TO CRITERION
(B-C) REGRESSION METHOD FOR A REGION WITH LOW CONDUCTIVITY	7-1
7.1.	DATA SET CHARACTERISTICS	7-1
7.1.1.	Ecoregion Description	7-1
7.1.2.	General Data Set Description	7-2
7.1.3.	EPA-Survey Data Set	7-4
7.1.4.	State Data Set (EPA Region 10)	7-6
7.1.5.	U.S. Geological Survey (USGS) Data Set	7-7
7.1.6.	Modeled Mean Base Flow Background Specific Conductivity (SC)	7-9
7.1.7.	Characterization of Ionic Matrices	7-11
7.1.7.1.	EPA-Survey Data Set Ionic Characteristics	7-11
7.1.7.2.	State Data Set Ionic Characteristics	7-11
7.1.7.3.	U.S. Geological Survey (USGS) Data Set Ionic
Characteristics	7-11
7.1.8.	Comparison of Background Specific Conductivity (SC) Estimates	7-12
7.2.	Calculation of the Criterion CONTINUOUS Concentration (CCC)	7-14
7.2.1.	Calculation of the Ecoregion 4 mean Hazardous Concentration (HCos)
from Background	7-14
7.2.2.	Calculation of the Lower 50% Prediction Limit	7-16
7.2.3.	Example Criterion Maximum Exposure Concentration	7-18
VI

-------
CONTENTS (continued)
7.3. EXAMPLE CRITERION CHARACTERIZATION FOR ECOREGION 4
BASED ON A BACKGROUND-TO-CRITERION MODEL	7-20
7.4. PROTECTION OF FEDERALLY-LISTED SPECIES AND OTHER
HIGHLY VALUED TAX A	7-21
8. REFERENCES	8-1
APPENDIX A: CASE STUDY I SUPPORTING MATERIALS	A-l
APPENDIX B: CASE STUDY II SUPPORTING MATERIALS	B-l
APPENDIX C: EXAMPLE CASE FOR ASSESSING BACKGROUND SPECIFIC
CONDUCTIVITY	C-l
APPENDIX D: DEVELOPMENT OF A BACKGROUND-TO-CRITERION
REGRESSION MODEL	D-l
APPENDIX E: COMBINED DATA SET FOR CASE STUDIES I AND II	E-l
APPENDIX F: CASE EXAMPLE USING AN ALTERNATIVE MEASURE OF
EXPOSURE (|IIC03 + SOr |)	F-l
APPENDIX G: CASE EXAMPLE USING AN ALTERNATIVE ASSESSMENT
ENDPOINT (SPECIES 01 MSI I)	G-l
vii

-------
LIST OF TABLES
2-1. Examples of ions associated with different anthropogenic sources	2-7
2-2. Comparison of methods to measure ionic concentration	2-12
2-3. Summary of assessment endpoints and measures of effect used in this field-based
method to derive a criterion continuous concentration (CCC) and criterion
maximum exposure concentration (CMEC) for specific conductivity	2-24
4-1. Summary statistics of the measured water-quality parameters used to derive the
example specific conductivity criteria in Ecoregion 69	4-5
4-2. Number of samples with reported genera and specific conductivity meeting
acceptance criteria for the Ecoregion 69 analysis	4-12
4-3. Summary data related to the calculation of the example criterion maximum
exposure concentration (CMEC) for Ecoregion 69	4-16
4-4. Description of survey data sets combined to form the EPA-survey data set used to
assess applicability of the example ecoregional criteria throughout Ecoregion 69	4-19
4-5. Summary of water quality parameters for Ecoregion 69 from the EPA-survey data
set excluding the sites in West Virginia	4-21
4-6.	Background specific conductivity estimates for Ecoregion 69	4-22
5-1.	Summary statistics of the measured water-quality parameters used to derive the
example specific conductivity criteria in Ecoregion 70	5-5
5-2. Number of samples with reported genera and specific conductivity meeting
data-inclusion acceptance criteria for the Ecoregion 70 analysis	5-12
5-3. Summary data related to the calculation of the example criterion maximum
exposure concentration (CMEC) for Ecoregion 70	5-16
5-4. Description of survey data sets combined to form the EPA-survey data set used to
assess applicability of example ecoregional criteria throughout Ecoregion 70	5-19
5-5. Summary of water quality parameters for Ecoregion 70 EPA-survey data set	5-21
5-6.	Background specific conductivity estimates for Ecoregion 70	5-21
6-1.	EPA and U.S. Geological Survey (USGS) chemistry data sets included in this
study	6-2
6-2. Summary of data for the example case for Northwestern Great Plains, Level III
Ecoregion 43 from EPA-survey samples	6-5
viii

-------
LIST OF TABLES (continued)
6-3. Summary of data for the example case for Northwestern Great, Level III
Ecoregion 43 from U.S. Geological Survey (USGS)	6-7
6-4. Summary of data for the example case for Northwestern Great Plains, Level III
Ecoregion 43 from the combined EPA-survey and U.S. Geological Survey
(USGS) data set	6-11
6-5.	Summary data related to the calculation of the example criterion maximum
exposure concentration (CMEC) for Ecoregion 43	6-17
7-1.	EPA and U.S. Geological Survey (USGS) chemistry data sets included in this
study	7-3
7-2. Summary of data for the example case for Cascades, Level III Ecoregion 4, from
EPA-survey samples	7-4
7-3. Summary of data for the example case for Cascades, Level III Ecoregion 4 from
State Data from Oregon and Washington	7-7
7-4. Summary of data for the example case for Cascades, Level III Ecoregion 4 from
the U.S. Geological Survey (USGS)	7-9
7-5. Summary of data for the example case for Cascades, Level III Ecoregion 4 from
the combined data set	7-14
7-6. Summary data related to the calculation of the example criterion maximum
exposure concentration (CMEC) for Ecoregion 4	7-19
IX

-------
LIST OF FIGURES
2-1. Conceptual model showing hypothesized relationships among selected sources of
ions and biotic responses to ionic stress by salt intolerant taxa (adapted from
Schofield and Ziegler, 2010)	2-9
2-2.	A species' (or genus') realized niche is defined by its lower and upper limits of
occurrence	2-15
3-1.	Example of a genus extirpation concentration distribution (XCD) depicting the
proportion of genera extirpated with increasing ionic concentration measured as
specific conductivity (SC)	3-2
3-2. Main steps in the derivation of a chronic specific conductivity (SC) criterion using
the extirpation concentration distribution (XCD) method	3-4
3-3. Examples of weighted cumulative distribution functions (CDFs) and the
associated 95fe centile extirpation concentration values	3-13
3-4. Examples of extirpation concentration (XC95S) for three genera listed as being
definitive (Neophylax), approximate (Amphinemura), and greater than
(Cambarus)	3-14
3-5. Diagram from EPA Benchmark Report depicting the process for estimating
uncertainty	3-17
3-6. Main steps in the derivation of a criterion maximum exposure concentration
(CMEC) based on field water chemistry data	3-21
3-7. Typical specific conductivity (SC) pattern of a stream with annual mean SC well
below 310 [j.S/cm	3-24
3-8. Correlation of specific conductivity and drainage area up to 17,986 km2,
Spearman's r = 0.25	3-35
3-9. Method for selecting a criterion continuous concentration (CCC) for a new area
within an ecoregion using minimally affected background	3-39
3-10. Empirical model of the 5th centile of a hazardous concentration (HC05) and
background specific conductivity (SC) estimated at the 25th centile for 23 distinct
ecoregions (24 data sets)	3-44
3-11.	A decision tree for calculating and applying a hazardous concentration of the
5th centile of a hazardous concentration (HC05)	3-47
4-1.	Ecoregion 69 extends from central Pennsylvania to northern Tennessee	4-4
x

-------
LIST OF FIGURES (continued)
4-2. Box plot showing seasonal variation of specific conductivity ([j,S/cm) from
probability-sampled sites from Watershed Assessment Branch database
(WABbase) 1997-2010	4-7
4-3. Box plot showing seasonal variation of specific conductivity ([j,S/cm) in the
reference streams from Watershed Assessment Branch database (WABbase)
1997-2010	4-8
4-4. Box plot showing seasonal variation of specific conductivity ([j,S/cm) from all
Ecoregion 69 sites from Watershed Assessment Branch database (WABbase)
1997-2010 used to develop the example criteria	4-9
4-5. Scatter plot of relationship between [CP] and ([HCO3 ] + [SO42 ]) concentrations
in streams in Ecoregion 69 data set from 1997-2010 with ionic measurements	4-11
4-6. Histograms of the frequencies of observed specific conductivity values in samples
from Ecoregion 69 sampled between 1997 and 2010	4-13
4-7. Example genus extirpation concentration distribution (XCD) for Ecoregion 69	4-14
4-8. The lower end of the example genus extirpation concentration distribution for
Ecoregion 69	4-15
4-9. Illustration of within site variability (residual standard deviation for each station)
along the specific conductivity gradient (station mean) in Ecoregion 69	4-17
4-10. Ecoregion 69 extends from central Pennsylvania to northern Tennessee	4-20
4-11. Relationship of the number of occurrences of a genus and the hazardous
concentration (HC05) based on Ecoregion 69 example Criterion-data set	4-24
4-12. The effect of the size of the data set used to model the hazardous concentration
(HC05) based on the Ecoregion 69 example Criterion-data set	4-25
4-13.	Cumulative distribution of the extirpation concentration (XC95) values for the
25% most salt-intolerant genera and 95% confidence intervals based on
1,000 extirpation concentration distribution (XCD) bootstrapping results	4-26
5-1.	Ecoregion 70 extends from central Pennsylvania to northern Tennessee	5-4
5-2. Box plot showing seasonal variation of specific conductivity ([j,S/cm) from
probability sites from Watershed Assessment Branch database (WABbase)
1997-2010	5-7
5-3. Box plot showing seasonal variation of specific conductivity ([j,S/cm) in the
reference streams of Ecoregions 70 from 1997 to 2010	5-8
xi

-------
LIST OF FIGURES (continued)
5-4. Box plot showing seasonal variation of specific conductivity ([j,S/cm) from all
Ecoregion 70 sites from Watershed Assessment Branch database (WABbase)
1997-2010 used to develop the example criteria	5-9
5-5. Scatter plot of relationship between [Cl~] and ([HCO3 ] + [SO42 ]) concentrations
in streams of Ecoregion 70 data set	5-11
5-6. Histograms of the frequencies of observed specific conductivity values in samples
from Ecoregion 70 sampled between 1997 and 2010	5-13
5-7. Example genus extirpation concentration distribution (XCD) for Ecoregion 70	5-14
5-8. The lower end of the example genus extirpation concentration distribution for
Ecoregion 70	5-15
5-9. Illustration of within site variability (residual standard deviation for each station)
along the specific conductivity gradient (station mean) in Ecoregion 70	5-17
5-10. Ecoregion 70 extends from southwestern Pennsylvania and southeastern Ohio into
Kentucky	5-19
5-11. Relationship of the number of occurrences of a genus on the hazardous
concentration (HC05) based on Ecoregion 70 example Criterion-data set	5-23
5-12. Adequacy of the size of the data set used to model the hazardous concentration
(HC05) based on the Ecoregion 70 example Criterion-data set	5-24
5-13.	95% confidence intervals (hatched oblique lines) for the lower portion of the
Ecoregion 70 genus extirpation concentration distribution (XCD)	5-25
6-1.	Sampling sites in the EPA-survey data set that were used to estimate minimally
affected background in Ecoregion 43	6-4
6-2. Sampling sites in the U.S. Geological Survey (USGS) data set in Ecoregion 43	6-6
6-3. Predicted mean natural base-flow specific conductivity in catchments of
Northwestern Great Plains, Ecoregion 43, using the Olson-Hawkins model	6-8
6-4. Box plots of specific conductivity (SC) distributions for EPA-survey, U.S.
Geological Survey (USGS), combined EPA-survey and USGS data sets and
predicted mean base-flow	6-10
6-5. Process and decision path case example for Ecoregion 43	6-13
6-6. Illustration of within site variability (residual standard deviation for each station)
along the specific conductivity gradient (station mean) in Ecoregion 43	6-16
xii

-------
LIST OF FIGURES (continued)
7-1. Sampling sites in the EPA survey data set that were used to estimate minimally
affected background in Ecoregion 4	7-5
7-2. Sampling sites in State data set that were used to estimate minimally affected
background in Ecoregion 4	7-6
7-3. Sampling sites in the U.S. Geological Survey (USGS) data set in Ecoregion 4	7-8
7-4. Predicted mean natural base-flow specific conductivity in catchments of the
Cascades, Ecoregion 4, using the Olson-Hawkins model	7-10
7-5. Box plots of specific conductivity (SC) distributions for EPA-survey, State,
U.S. Geological Survey (USGS), and combined data sets, and the predicted base
flow SC for all stream segments	7-13
7-6. Process and decision path case example for Ecoregion 4	7-15
7-7. Illustration of within site variability (residual standard deviation for each station)
along the specific conductivity gradient (station means) in Ecoregion 4	7-19
Xlll

-------
LIST OF ABBREVIATIONS AND ACRONYMS
B-C
b ackground-to-criteri on
CCC
criterion continuous concentration
CDF
cumulative distribution function
CI
confidence interval
CM EC
criterion maximum exposure concentration
DO
dissolved oxygen
EMAP
Environmental Monitoring and Assessment Program
EPA
U.S. Environmental Protection Agency
GAM
generalized additive model
HCx
hazardous concentration of the "x" centile of a taxonomic sensitivity distribution
LOWES S
Locally Weighted Scatterplot Smoothing
MAHA
Mid-Atlantic Highland Assessment
MAIA
Mid-Atlantic Integrated Assessment
NAPAP
National Acid Precipitation Assessment Program
NRSA
National Rivers and Streams Assessment
NWS A
National Wadeable Streams Assessment
PL
prediction limit
QA/QC
quality assurance/quality control
RBP
rapid bioassessment protocol
R-EMAP
Regional Environmental Monitoring and Assessment Program
S
Siemens
SAB
Science Advisory Board
SC
specific conductivity
TDS
total dissolved solids
TMDL
total maximum daily load
TSS
total suspended solids
USGS
U.S. Geological Survey
WABbase
Watershed Assessment Branch database
WDE
Washington Department of Ecology
WSA
Wadeable Streams Assessment
WVDEP
West Virginia Department of Environmental Protection
XCx
extirpation concentration affecting "x" percentage of individuals of a taxon
XCD
extirpation concentration distribution
xiv

-------
NOTICES
This public review draft document has undergone two contractor-led external peer
reviews as well as a review process within the U.S. Environmental Protection Agency (EPA).
Final review by EPA's Office of Science and Technology, Health and Ecological Criteria
Division, has been completed and the document has been approved for publication.
This document provides draft methods to assist states and tribes in the development of
water quality criteria and other tools to protect aquatic life from effects of elevated ionic
concentration as measured by specific conductivity (SC)1 in flowing waters. States and tribes
planning to develop water quality criteria for SC may consider using alternative, scientifically
defensible methods. While this document reflects EPA's assessment of the best available
science for identifying ambient concentrations of SC in flowing waters that protect aquatic life, it
is not a regulation and does not impose legally binding requirements on EPA, states, tribes, or
the regulated community, and might not apply to a particular situation based upon the
circumstances. EPA may change this document in the future.
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use. This document can be downloaded from:
https://www.epa.gov/wqc/aquatic-life-ambient-water-quality-criteria.
Cover Photo:
Used by permission, from Randall Sanger Photography. Photo of New River, West Virginia.
'This document uses conductivity as a measure of ionic concentration rather than as description of an electrical
property of water. As ionic concentration increases, conductivity increases. The terms specific conductivity and
specific conductance are often used synonymously in the open literature indicating normalization or measurement at
25°C. Conductivity is a property of water expressed in units of micro-Siemens per centimeter (|iS/cm).
Conductance of a sample or electrical component is measured as Siemens (S). All measurements in this document
refer to specific conductivity, (iS/cm at 25°C.
xv

-------
FOREWORD
This document, Draft Field-based Methods for Developing Aquatic Life Criteria for
Specific Conductivity, provides states and tribes with methods that may be used to develop
criteria to protect aquatic life from effects of elevated ionic concentration as measured by
specific conductivity (SC) in flowing waters. The EPA tailored these methods to enable
derivation of specific conductivity criteria on the scale of Level III ecoregions (Omernik, 1995,
1987) in order to account for natural differences in background ionic concentrations among
ecoregions. There are 85 Level III ecoregions in the contiguous United States. Each of the
states in the contiguous United States contains 1 to 12 Level III ecoregions within their political
boundaries. The EPA is also providing several case studies to illustrate how these draft methods
may be applied to different ecoregions with varying background ionic concentrations. The EPA
may change the field-based methods and/or provide additional case studies in the future as new
scientific information becomes available.
This document is nonregulatory and provides only a scientific assessment of ecological
effects. It does not establish or affect legal rights or obligations. It does not establish a binding
norm and cannot be finally determinative of the issues addressed. Agency decisions in any
particular situation will be made by applying the Clean Water Act and EPA regulations on the
basis of specific facts presented and scientific information then available.
Elizabeth Southerland
Director
Office of Science and Technology
xvi

-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
Prepared by:
Colleen Flaherty and Rachael Novak
U.S. EPA, Office of Water, Office of Science and Technology
Health and Ecological Criteria Division
Washington, DC
Susan Cormier, Ph.D., Glenn Suter II, Ph.D., and Michael Griffith, Ph.D.
U.S. EPA, Office of Research and Development
National Center for Environmental Assessment
Cincinnati, OH
Contract Support:
Lei Zheng Ph.D.
Tetra Tech, Inc.
Owings Mills, MD
Dennis Mclntyre and Keith Taulbee
Great Lakes Environmental Center
Columbus, OH
Other contributors: Diane Allen, Jerome Diamond, Jennifer Flippin, Jeroen Gerritsen, Andrew
Hamilton, Benjamin Jessup, Erik Leppo, James T. Robbiani, Jenifer Stamp, Christopher
Wharton, Samuel P. Wilkes
Tetra Tech, Inc. (Owings Mills, MD)
EPA Review:
The draft document benefitted from the insights of an internal EPA technical workgroup, in
particular the following EPA staff:
Office of Water: Ruth Chemerys, Rosaura Conde, Charles Delos, Kathryn Gallagher,
Thomas Gardner, Ross Gerredien, Lisa Huff, Chris Hunter, Matthew Klasen, Eric
Monschein, Brian Topping, Scott Wilson, Marcus Zobrist
Office of Research and Development. Robert Cantilli, Russell Erickson, Alan Herlihy,
Russ Hockett, Dale Hoff, Scott Leibowitz, David Mount, Cynthia Roberts, Charles
Stephan, Michael Troyer, and Ryan Hill (Oak Ridge Institute for Science and Education
Research participant)
Office of Policy. Bridgid Curry, Ann Johnson, Sharon Cooperstein
Office of General Counsel. Michael G. Lee, Alexis Wade
Region 3: Francisco Cruz, Gregory Pond, Margaret Passmore, Louis Reynolds
xvii

-------
Region 4: Larinda Tervelt, Kip Tyler, Leah Ettema, Katherine Snyder
Region 5: Edward Hammer, Peter Jackson, Krista McKim, Robert Pepin, Kerryann
Weaver, Candice Bauer
Region 8: Tonya Fish, Lareina Guenzel, Tina Laidlaw, Peter Brumm, George Parrish
Region 10: Gretchen Hayslip
Document Formatting and Editing.
Contract management: Bette Zwayer, EPA, ORD.
TSS Document Processing (formatting): Kathleen Bland, Highlight Technologies and Debbie
Kleiser, Highlight Technologies
TSS Technical Editing; Tom Schaffner, CACI, and Lisa Walker, CACI
TIU Reference Management and Quality Control; Linda Tackett, CACI
Independent External Review:
David Buchwalter, Ph.D.
Department of Biological Sciences
North Carolina State University
Raleigh, NC
Yong Cao, Ph.D.
Illinois Natural History Survey
University of Illinois at Urbana-Champaign
Champaign, IL
Bruce K. Hope, Ph.D.
Independent Consultant
Medford, OR
Marian R.L. Maas, Ph.D.
Independent Consultant
Bellevue, NE
Raymond P. Morgan II, Ph.D.
University of Maryland Center for
Environmental Science
Appalachian Laboratory
Frostburg, MD
Edward T. Rankin, M.S.
Midwest Biodiversity Institute
Columbus, OH
Carl E. Zipper, Ph.D.
Department of Crop and Soil Environmental
Sciences
Virginia Polytechnic Institute and State
University
Blacksburg, VA
xviii

-------
EXECUTIVE SUMMARY
This document describes draft methods that states and tribes may use to derive
field-based ecoregional ambient aquatic life criteria for specific conductivity (SC), a
measurement of the concentration of ions, in flowing waters. The document also provides four
case studies to illustrate how these draft field-based methods may be used to develop criteria in
ecoregions with different background ionic concentrations measured as SC and to demonstrate
how to assess the applicability of criteria developed for one ecoregion to a different ecoregion.
The case studies use field data to demonstrate how to apply the methods described in this
document to derive example criteria for SC for flowing waters dominated by calcium,
magnesium, sulfate, and bicarbonate ions but not for flowing waters dominated by chloride salts.
Elevated ionic concentration measured as SC has been shown to impact aquatic life in a range of
freshwater resources. Different mixtures of ions that increase SC are associated with multiple
anthropogenic sources, including discharges from wastewater treatment facilities, groundwater
recharge affected by climate change, surface mining, oil and gas exploration, runoff from urban
areas, and discharges of agricultural irrigation return waters, among others.
The EPA relied on its Guidelines for Deriving Numerical National Water Quality
Criteria for the Protection of Aquatic Organisms and Their Uses (1985) (EPA/822/R-85/100)
and A Field-Based Aquatic Life Benchmark for Conductivity in Central Appalachian Streams
(hereafter referred to as the "EPA Benchmark Report") (EPA/600/R-10/023F), among other
documents, to develop the draft field-based method for SC. In the EPA Benchmark Report, EPA
used a field data set to estimate a numeric SC benchmark for Appalachian streams. The EPA
validated the method and the benchmark using an independent data set. In 2011, internal and
external reviewers, including EPA's Science Advisory Board (SAB) (U.S. EPA, 201 lc),
favorably reviewed the analyses and method. This current document uses that same method to
estimate a protective criterion continuous concentration (CCC) for chronic (long-term) exposures
as well as additional methods to estimate a maximum exposure concentration protective of acute
toxicity. This document also provides recommendations for SC criterion duration and frequency.
The EPA typically relies on laboratory toxicity test data for surrogate species as defined
in the Agency's Guidelines for Deriving Numerical National Water Quality Criteria for the
Protection of Aquatic Organisms and Their Uses (U.S. EPA, 1985) for aquatic life criteria
xix

-------
development. The draft field-based methods used here were adapted to be consistent with the
intent of the Agency's traditional approach to derive aquatic life criteria (U.S. EPA, 1985). The
draft field-based methods rely on geographically referenced, paired observations of SC and the
presence and absence or abundance of freshwater benthic macroinvertebrate genera from
wadeable perennial streams. The case studies that are included to illustrate the method are based
on more than 4,000 paired biological (macroinvertebrate) and chemical (SC) field samples from
more than 3,000 stations over a 15-year period (1996-2010). An analysis of data for fish from a
composite of case study ecoregions demonstrates that the example criteria based on
macroinvertebrates are also protective of fish.
For this draft field-based method, the valued resource is the aquatic community. The
ecological entities defining the assessment endpoints are macroinvertebrate genera and the
measure of effect is extirpation, or effective absence of such genera from a site (the desired
attribute is occurrence). Two relationships are derived: one for each macroinvertebrate genus
and one for the overall aquatic community. First, a weighted cumulative distribution function
(CDF) is developed for each genus to determine the genus extirpation concentration (XC95 or
95th centile of the distribution of the occurrences of a genus), the level of exposure above which
a macroinvertebrate genus is effectively absent from water bodies in a region or other study area
(U.S. EPA, 201 la, 2003). That is, the probability is 0.05 that an observation of a genus would
occur above its XC95 SC value. Second, the HC05 (hazard concentration 5th centile) is developed
using a genus-level extirpation concentration distribution (XCD) for the community from the
aggregation of the XC95 values. This effect threshold is consistent with the intent of EPA's
guidelines for aquatic life criteria development (U.S. EPA, 1985), which are designed to protect
aquatic animal species (i.e., 95%) in a community.
The HCos is a chronic-duration endpoint and used for derivation of a CCC because it is
derived from biological field data that include exposure over whole life cycles and multiple
generations of the resident biota. A criterion maximum exposure concentration (CMEC), a level
of protection from acutely toxic exposures, is also derived based on stream water chemistry data.
The CMEC is estimated at the 90th centile of observations at sites with water chemistry regimes
meeting the CCC. The CMEC is the maximum SC level that may occur for a short duration and
be protective of 95% of macroinvertebrate genera. Both of these distinct expressions of the
example SC criteria would need to be met in order to adequately protect aquatic life.
xx

-------
The type of model used in this draft method, a genus-level XCD, describes how genera in
biotic communities in general respond to a stressor (e.g., an ionic mixture dominated by sulfate
and bicarbonate salts). This method is based on a distribution of extirpation concentrations and
is called the XCD method to distinguish it from other field-based methods. Like the surrogate
aquatic taxa that form the minimum data set for laboratory-based aquatic life criteria, the
macroinvertebrate taxa included in the case studies are surrogate taxa that represent a potentially
exposed aquatic community (U.S. EPA, 1985).
xxi

-------
GLOSSARY
Assessment endpoint—An explicit expression of the actual environmental value that is to be
protected, operationally defined by an ecological entity and its attribute or characteristics.
An assessment endpoint may be identified at any level of organization (e.g., organism,
population, community).
Assemblage, stream—A taxonomic or sampled subset of a community as may be collected from
a stream.
Background specific conductivity—The specific conductivity (SC) in streams in a region that
occurs naturally and not as the result of human activity. Background may also be
characterized as a population of minimally affected sites or low SC sites using a weight
of evidence.
Benchmark—A dose or concentration of a pollutant that, if exceeded, is expected to produce an
adverse effect (called the benchmark response) in one or more assessment endpoints,
signifying a decline in water quality or human health.
Bootstrapping—A statistical technique of repeated random sampling from a data set that is often
used in environmental studies to estimate confidence and prediction limits of a parameter.
Box plot—A depiction of the 25th, 50111, and 75111 quantiles of a distribution as a rectangle with a
central line. The two standard deviation range is depicted as "whiskers" extending from
the box. Data beyond two standard deviations are indicated by individual circles or dots
beyond the whiskers.
Catchment area—The spatial extent of the surrounding landscape that drains into a particular
river, stream, or other waterbody.
Chorionic covering—The outermost casing or membranous covering of the egg of various
invertebrates.
Community—The full complement of interacting organisms within a defined area of an
ecosystem.
Conductivity, specific (or specific electrical conductivity)—A measure of ionic concentration
based on the electrical property of water and dissolved ions. As ionic concentration
increases, conductivity increases. Standardized measurements in this document refer to
specific conductivity, [j,S/cm (also seen as: (j,mho/cm) at 25°C.
Conductance, specific—Conductance is the inverse of resistance for a particular sample
expressed as Seimens (S) usually at 25°C. In the literature, it is sometimes used
synonymously with specific conductivity, but to avoid confusion, the term conductance is
not used in this document.
Confounder—An extraneous variable that correlates with both the dependent and independent
variable. The presence of confounders can interfere with the ability to characterize a
causal relationship.
Criterion continuous concentration (CCC)—An estimate of the highest concentration of a
material in surface water to which an aquatic community can be exposed indefinitely
without resulting in an unacceptable effect.
Criterion maximum exposure concentration (CMEC)—An estimate of the maximum
concentration of a material in surface water to which an aquatic community can be
exposed for a short time without resulting in an unacceptable effect. In this document,
the CMEC is estimated at the 90th centile of specific conductivity observations that
contribute to the annual CCC.
xxii

-------
Cumulative distribution function (CDF)—The probabilities that a random variable with a given
probability distribution will be found at a value less than or equal to x. Weighted CDFs
are used to estimate extirpation concentrations of individual genera or species and
unweighted CDFs to estimate a SC level that is expected to extirpate 5% of aquatic
invertebrate genera.
Ephemeral stream—A stream that flows briefly only in direct response to local precipitation, and
whose channel is above the local groundwater table at all times.
Extirpation—The depletion of a population of a species or genus to the point that it is no longer a
viable resource or is unlikely to fulfill its function in the ecosystem.
Extirpation concentration—The level above which a genus is effectively absent from its normal
habitat. The threshold for extirpation is operationally defined by the level below which
95% of the observations of the genus occur.
Extrapolation—The process of extending the applicability of a model beyond the measured range
of the original data set from which the model was derived.
Flowing waters—Inland waters with a unidirectional flow including permanent, intermittent and
ephemeral streams.
Generalized additive model—A nonparametric, likelihood-based local regression model that
replaces the linear function of a generalized linear model with a locally smoothed
additive function.
Hazardous concentration—A concentration threshold that is hazardous for a proportion of taxa.
In this document, it is the concentration that is hazardous to 5% of genera calculated as
the 5th centile of a taxonomic extirpation concentration distribution.
Intermittent stream—A stream that flows continuously for only part of the time. During low
flow there may be dry reaches alternating with wetted, nonflowing reaches. The stream
bed may lie below the local groundwater table for at least part of the year.
Interpolate—Process of estimating an unknown value that lies between known values.
Ionic composition—The specific ions dissolved in water. In this document, the ionic
composition is used to distinguish water dominated by chloride salts from those
dominated by bicarbonate and sulfate salts.
Ionic mixture—An undefined or defined blend of dissolved ions. In this document, the example
case studies refer to the most common mixture of ions contaminating U.S. streams,
specifically those dominated by calcium, magnesium, sulfate, and bicarbonate ions.
Ionic regulation—The passive and active physiological processes that maintain the ionic
composition, pH, and water content of tissues that is necessary for life.
Least disturbed condition—the best available physical, chemical, and biological habitat
conditions given today's state of the landscape or the least disturbed by human activities
(Stoddard et al., 2006). Contrast with "minimally affected condition."
Major ions—The most common contributors to ionic concentration in surface waters, consisting
of the following cations: Ca2+, Mg2+, Na+, K+; and anions: HCO3 , CO32 , SO42 , CP.
Measure of effect—A measurable ecological characteristic that is related to the valued
characteristic chosen as the assessment endpoint and is a measure of biological effects
(e.g., survival, reproduction, growth). In this document it is the presence/absence of
macroinvertebrate genera along a specific conductivity gradient.
Measure of exposure—A measured or estimated characteristic that is used to characterize the
level of exposure to the stressor. In this document, the measure of ionic exposure is
specific conductivity.

-------
Minimally affected condition—The physical, chemical, and biological habitat found in the
absence of significant human disturbance (Stoddard et al., 2006). Contrast with "least
disturbed condition."
Osmoregulation—The physiological control of water content of an organism's tissues to maintain
fluid and electrolyte balance within a cell or organism relative to the surrounding
environment.
Perennial stream—A stream with continuous surface or shallow interstitial flow year-round, and
whose stream bed intersects the local groundwater table throughout the year. Also
referred to as a permanent stream.
Produced water—Waters that are produced by oil and gas development, mine dewatering, and
related activities (e.g., coal bed methane mining, hydraulic fracturing).
Reference site—Sampling locations that have been identified as minimally affected or least
disturbed based on land use, habitat, and water quality characteristics other than specific
conductivity.
Salinity—The amount of salts dissolved in water. Traditionally expressed as parts per thousand
(%o) or grams of salt per kilogram of water.
Sensitivity analysis—A process that involves changing input values of a model in various ways
to see the effect on the output value. The main goal of sensitivity analysis is to gain
insight into which assumptions are most critical for model building.
Total dissolved solids (TDS)—A measure of the combined content of all inorganic and organic
substances dissolved in water, conventionally expressed as mg/L and operationally
defined as those solids that pass through a filter, typically 0.45 [j.m.
Univoltine—An organism having one brood or generation per year.
Validation—Confirmation of the quality of a model and its results, typically by applying an
independent data set.
Valley fill—A headwater valley filled with mining overburden. This practice usually occurs in
steep terrain where there are limited disposal alternatives.
Verification—Demonstrating the accuracy of measurements or calculations.
xxiv

-------
1. INTRODUCTION AND BACKGROUND
This document describes a set of draft methods that states and authorized tribes may use
to derive field-based ecoregional ambient aquatic life criteria for ionic mixtures measured as
specific conductivity (SC), a measurement of ionic concentration. Four case studies illustrate
how these draft methods may be applied to develop such criteria in different ecoregions with
different background SC and data sets. The case studies illustrate how these methods may be
used to develop criteria applicable to flowing waters dominated by sulfate and bicarbonate salts.
Chloride constitutes less than half of the total anions in the case examples. Although the
methods may be appropriate for use with other ionic mixtures, the example criteria generally are
not appropriate for waters with different ionic compositions (e.g., waters dominated primarily by
sodium chloride).
Among the documents the U.S. Environmental Protection Agency (EPA) relied upon to
develop the draft field-based method for SC are EPA's Guidelines for Deriving Numerical
National Water Quality Criteria for the Protection of Aquatic Organisms and Their Uses
(U.S. EPA, 1985) and A Field-Based Aquatic Life Benchmark for Conductivity in Central
Appalachian Streams (hereafter referred to as the "AYM Benchmark Repor f) (U.S. EPA, 201 la).
The EPA used an extensive field data set in the EPA Benchmark Report to estimate a numeric
SC benchmark. The EPA validated the method and benchmark using an independent data set.
The EPA Benchmark Report provides details on the approach, as well as a causal analysis of the
stressor-response relationship and a confounder analysis that explored the potential influence of
habitat, water quality factors, other pollutants, and other factors. Internal and external reviewers,
including EPA's Science Advisory Board (SAB), reviewed the primary method and derivation of
the SC benchmark and validation exercises in 2011 (U.S. EPA, 201 lc). Subsequently, the
method and results of its application were published (Cormier and Suter, 2013a, b; Cormier
et al., 2013a, b, c; Suter and Cormier, 2013). This current draft document uses that method as
well as additional methods to estimate a protective maximum exposure concentration, duration,
and frequency. It also presents a draft method for assessing applicability of field-based SC
criteria developed in one geographic area to another area. In 2014 and 2015, panels of five
external experts (selected independently by an EPA contractor) reviewed these additional draft
methods.
1-1

-------
These draft field-based methods may be used to develop SC criteria on the scale of
Level III ecoregions (Omernik, 1995, 1987) in order to take into account natural ecoregional
differences in background SC. In some areas, it may be appropriate to derive criteria at a
different scale because background conductivity or ionic composition varies significantly across
a Level III ecoregion (see Section 6 for an example). There are 85 Level III ecoregions in the
continental United States (Omernik, 1995, 1987). SC tends to be low in most eastern and
western montane ecoregions (25th centiles of SC <200 (j,S/cm), intermediate in the midcontinent
(200-600 (j,S/cm), and very high in arid areas (>600 (j,S/cm) (Griffith, 2014). States and tribes
may use this method to derive ecoregional criteria for SC at a level that protects 95% of resident
macroinvertebrate genera based on field sampling data from a set of sites within the ecoregion or
from another ecoregion, when applicable.
The EPA typically relies on laboratory toxicity test data as defined in the Agency's
Guidelines for Deriving Numerical National Water Quality Criteria for the Protection of Aquatic
Organisms and Their Uses (U.S. EPA, 1985). EPA designed the draft field-based methods
described herein to be consistent with the intent of the Agency's traditional approach used to
derive aquatic life criteria (U.S. EPA, 1985). Like the Agency's traditional approach, criteria
derivation through field-based methods can capture characteristics of the stressor and the
ecosystems potentially at risk (e.g., stressor occurrence and distribution, stressor-response
relationships).
The structure of this draft document, Field-based Methods for Developing Aquatic Life
Criteria for Specific Conductivity, is consistent with the EPA's Guidelines for Ecological Risk
Assessment (U.S. EPA, 1998a; Suter and Cormier, 2008). The assessment begins with a
planning phase, termed Problem Formulation (see Section 2), in which the stressor of concern is
identified, its presence in the environment and potential impacts are described, and assessment
endpoints (i.e., specific ecological entities and attributes to be protected and the level of
protection to be achieved) are identified. In the case studies, the stressor is a mixture of ions in
the form of dissolved bicarbonate and sulfate salts, measured as specific SC, in the field. The
endpoint populations are aquatic benthic macroinvertebrates and the measure of effect is
extirpation not to exceed 5% of genera. Section 2 serves as the Problem Formulation in general
and for all four case studies.
1-2

-------
In this draft document, the Analysis Plan (see Section 3), which is the last step in
Problem Formulation, is included as a separate stand-alone section. The Analysis Plan describes
three methods, (1) a field-based method that states may use to directly derive field-based aquatic
life criteria for SC (the extirpation concentration distribution [XCD] method), (2) a regression
model that can be used to derive criteria from minimally affected background (the
background-to-criterion [B-C] model method), and (3) a method to assess the geographic
applicability (extent) of the criteria using a weight-of-evidence approach. Section 3 serves as the
Analysis Plan for this draft method in general and for all four of the case studies that follow in
the Case Study Analysis sections. Each of the methods considers the causal relationship between
exposure to major aqueous ions and the response of macroinvertebrates.
Next, in the Case Study Analysis sections (see Sections 4 and 5), the application of the
draft XCD method is illustrated by deriving example SC criteria for different ecoregions with
ecoregion-specific data sets. These sections describe magnitude, frequency, and duration as well
as factors characterizing geographic range (see Case Studies I and II, Sections 4 and 5). Two
other case studies demonstrate how to use the B-C regression method that predicts criteria from
minimally affected background (see Case Studies III and IV, Sections 6 and 7). In these case
studies, there are several factors relevant to determining geographic applicability (spatial extent
of the criteria); among the most important are background SC and the composition of the ionic
mixture present (ions of bicarbonate and sulfate salts).
Appendices A and B provide supporting materials, including assessments of potential
confounding factors, and plots and effect levels for all genera represented in ecoregional XCDs
used in the development of the Case Studies I and II (see Appendices A for Case Study I and B
for Case Study II). Appendix C discusses the characterization of background SC and the
seasonal regime of a region (a condition assessment) and includes a specific example for Case
Study II. Appendix D provides the derivation of a B-C regression model that uses minimally
affected background SC to calculate a SC criterion that is useful for areas lacking sufficient data
to use the XCD method (see application of this model in Case Studies III and IV). Appendix E
provides extirpation concentration (XC95) values for the combined data sets used for Case
Studies I and II. Appendix F provides results using an alternate measure of the ionic mixture,
sulfate plus bicarbonate (as mg/L). Appendix G provides an analysis that shows that some fish
1-3

-------
in streams are intolerant of high ionic concentrations and that fish are protected by criteria
derived by applying the XCD method to benthic invertebrate data.
Data quality reviews of project data sets were conducted to ensure that the data used and
the results of the analyses are accurate and complete. When invalid or incorrect data were
identified, these data were either corrected or excluded from analyses. Methods for data
extraction, data management, model development, and quality assurance/quality control
(QA/QC) for this project are described in the Quality Assurance Project Plan, prepared by Tetra
Tech, Inc. 2014. Validation and other QA analyses are described as each model or case study are
also presented.
1-4

-------
2. PROBLEM FORMULATION
This section serves as the Problem Formulation for the XCD method in general and for
the case studies, which are presented in Sections 4, 5, 6, and 7. Problem Formulation begins
with identification of the problem (see Section 2.1), the stressor of concern and its sources (see
Section 2.2), and a description of how it can be measured (see Section 2.3). In the case
examples, the stressor is a mixture of ions in a form dominated by bicarbonate and sulfate salts,
measured using SC. The nature of effects (see Section 2.4), and mechanisms and modes of
action are described (see Section 2.5). The assessment endpoints and measures of effect are
described (see Section 2.6). The organisms are freshwater benthic macroinvertebrates and the
measurement of effect is extirpation of 5% of genera. Extirpation is the depletion of an
assessment population of a species or genus (in this case, it is the population in a stream) to the
point that it is no longer a viable resource or is unlikely to fulfill its function in the ecosystem
(U.S. EPA, 2003). Specifically, this effect threshold is defined in this document as the ionic
concentration below which 95% of the observations of the genus occur, representing the extreme
of an organism's tolerance to an ionic mixture. In the case studies, the ionic mixture as
measured by SC is dominated by sulfate and bicarbonate salts, with either calcium and
magnesium or sodium and potassium as the cations (U.S. EPA, 201 la). This effect threshold is
consistent with the intent of EPA's guidelines for aquatic life criteria development (U.S. EPA,
1985), which are designed to protect aquatic animal species (i.e., 95%) in a community. The
Problem Formulation section concludes with the rationale for selection of a field-based method
for derivation of criteria for the ionic mixture (see Section 2.7).
2.1. PROBLEM IDENTIFICATION
Stress from elevated ionic concentration, measured as specific SC, has been shown to
cause significant adverse effects on a range of freshwater ecosystems across the Nation
(e.g., Caneda-Argiielles, et al., 2013; Higgins and Wilde, 2005; Kaushal et al., 2013, 2005; Pond
et al., 2008; U.S. EPA, 201 la). The sources of ions in surface waters may be natural, reflecting
soils and geology, or anthropogenic. The two most common ionic mixtures in streams are those
dominated by either chloride anions (CP) or those dominated by bicarbonate (HCO3 ) plus
sulfate (SO42 ) anions based on mass (Hem, 1985; Griffith, 2014). The field-based methods are
2-1

-------
illustrated using case examples with flowing waters with ionic mixtures dominated by HCO3
plus SO42Based on mass, CP constitutes less than half of the total anions in the case examples.
2.2. STRESSOR OF CONCERN—SALTS
Ionic stress has been implicated as a cause of biological impairment in aquatic systems
throughout the United States (e.g., Findlay and Kelly, 2011; Farag, and Harper, 2012; Dunlop
et al., 2015; Boelter, et al., 1992; Higgins and Wilde, 2005; Johnson et al., 2013; Karatayev
et al., 2012; Kaushal et al., 2013, 2005; Fritz et al., 2010; Gerritsen et al., 2010; Palmer et al.,
2010; Lindberg et al., 2011; Merriam et al., 2011; Pond et al., 2008, Pond, 2010; U.S. EPA,
201 la,b; Bernhardt et al., 2012; Cormier et al., 2013b; Timpano et al., 2011; Zhao et al., 2016).
Nationally, sources of salts can be natural from rock formations and soils or can be associated
with human activities and may be exacerbated by changes in climate. Sources include coastal
salt water intrusion, irrigation, combustion wastes, resource exploration and extraction,
demineralization of concrete, runoff from urban areas, inputs from deicing roads, and sewage
and industrial waste (Ziegler et al., 2010; Caneda-Argiielles, 2013) (see Table 2-1).
Furthermore, salts from different sources have different ionic compositions. For example,
marine evaporite deposits are dominated by NaCl whereas weathering of minerals such as
limestone and dolomite produce Ca2+, Mg2+, and HCO3 salts (Hem, 1985).
Consistent with the EPA Benchmark Report (U.S. EPA, 201 la), these draft field-based
methods may be applied for any waters with a defined ionic composition because the toxicity to
aquatic organisms is dependent on the ionic composition of the solution (Mount et al., 1997;
Mount et al., 2016; Erickson et al., 2016; Zalizniak et al., 2006; van Dam et al., 2010; Dunlop
et al., 2005; Soucek and Kennedy, 2005; Bradley, 2009; Evans, 2008a,b; Nelson and Cox, 2005,
Johnson et al., 2015). Aquatic organisms are adapted to different ionic regimes and have
different tolerances to changes in ionic concentration and composition (Remane, 1971; Bradley,
2009). Although certain species, particularly of fish and Crustacea, have life histories and
ionoregulatory adaptations that facilitate movement across a salinity gradient (Belli et al., 2009),
most groups have distinct lineages of orders and families that are limited to either freshwater or
marine environments (Remane, 1971; Berra, 2007). Outside of the physiological tolerance of a
species, the toxicity of salts interferes with ionic regulation, osmoregulation, and acid-base
balance (Bradley, 2009; Nelson and Cox, 2005).
2-2

-------
Because toxicities of ions differ and because the example criteria are derived with data
for streams where Ca2+ plus Mg2+, and HCO3 plus SO42 (i.e., not Na+ and CP) dominate the
ionic composition on a mass basis, the case example criteria are not recommended for locations
where CP concentrations are greater than the combined concentrations of HCO3 plus SO42 .
However, the XCD method could be used to derive criteria for other ionic mixtures, including
locations dominated by CP.
Application of this XCD method relies on the availability of paired chemical and
biological samples taken from waters with similar ionic composition (e.g., sulfate- and
bicarbonate-dominated). The sites included in the data sets are screened based on ionic
composition (e.g., chloride-dominated sites are removed from the data set in the case examples).
However, removing them did not appreciably change the results in the case examples because
there were so few sites that were chloride dominant.
2.2.1. Sources of Ions
Most fresh waters in the United States exhibit rock dominance (i.e., ion concentrations
characteristic of natural weathering of minerals in the catchment) (Gibbs, 1970; Stallard and
Edmond, 1987; Anning and Flynn, 2014), and the anion signature of these waters is usually
dominated by HCO3 plus SO42 (Wetzel, 2001; Griffith, 2014). SC tends to be low in
mountainous and forested ecoregions (25th centiles of SC -50-200 (j,S/cm) and higher in more
arid ecoregions (Griffith, 2014; Anning and Flynn, 2014). Nationally, the dominant cation
combination is calcium (Ca2+) plus magnesium (Mg2+) and the dominant anions combination is
bicarbonate (HCO3 ) plus sulfate (SO42 ) (Griffith, 2014). Exposure of soils and geologic
formations to weathering is a natural source of ions (Olson and Hawkins, 2012; Hem, 1985;
Pond, 2004; U.S. EPA, 201 lb). Factors such as rock texture and porosity, regional structural
geology, the degree of fissuring (or fracturing), exposure time, and other factors may influence
the composition of water flowing over and percolating through rocks (Hem, 1985). Igneous and
metamorphic rocks do not increase the ionic concentration of water flowing over them as much
as sedimentary rocks because they are generally more resistant to weathering (Anning et al.,
2007). Carbonaceous sedimentary rocks, such as limestone (CaCCb) and dolomite
(CaMgfCCb]), are sources of Ca2+, HCO3 and Mg2+, while other sedimentary rocks such as
those containing gypsum (CaSC>4 2H2O) and anhydrite (CaSC>4) can be natural sources of SO42 ,
2-3

-------
particularly in arid regions (Hem, 1985). Sedimentary rocks and salt deposits associated with
evaporation, such as ancient sea-beds, may contain high levels of Na+ and CP. Natural geologic
variability among neighboring watersheds may result in differences in ionic concentration of
associated streams. The ionic concentration of surface waters may increase naturally due to
evapotranspiration, evaporation, or recharge from groundwater with higher ionic concentrations.
Precipitation (e.g., rain or snow melt) can also affect ionic concentration. SC increases
during episodes of below-normal surface flow and decreases during periods of above-normal
surface flow. Seasonal patterns can vary greatly with regional climate, with low SC associated
with spring rain or during summer from snow-melt. Aerial deposition of wet and dry SO42
strips soils of Ca2+ and Mg2+ and thus directly and indirectly increases SC (Krug and Frink,
1983; Kaushal et al., 2013). Near ocean coastlines, rain and dry deposition may contain more
Crfrom entrainment of aerosols from seawater (Griffith, 2014). Pure water has low SC, due to
low concentrations of ions in solution. Surface and ground waters have a wide range of SC,
from <50 microsiemens per centimeter ([j,S/cm), where water quality is dominated by rainfall and
rocks are resistant to weathering, to over 200,000 [j,S/cm for brines (Hem, 1985).
Anthropogenic sources of ions can contribute to changes in both the ionic composition
and concentration in freshwater resources. Human activities can increase the ionic concentration
of natural waters either directly (e.g., by introducing new ions to freshwater systems) or
indirectly (e.g., by changing land use to those that increase delivery of ions to freshwater systems
and reduce freshwater input and recharge). For example, industrial, residential, and commercial
activities may discharge ion-rich waters to surface water. Reservoirs increase evaporation, thus
concentrating ions. Ionic concentration in freshwater systems can also increase as the result of
discharges of brines and wastes from combustion effluents or mines, and runoff from treating
pavements for icy conditions. Mining practices remove overlying vegetation and use explosives
to break up underlying rock, leading to increased ionic leaching from mine overburden as well as
from oxidation of exposed minerals such as pyrite (Johnson and Johnson, 2015; Bernhardt and
Palmer, 2011; Fritz et al., 2010; Lindberg et al., 2011; Merriam et al., 2011; Palmer et al., 2010;
Pond, 2010; Pond et al., 2008; Sams and Beer, 2000). Some mining practices deposit loosely
packed spoils comprised of crushed rock overburden into valley fills, where both chemical
leaching due to rainfall and direct transport of ions bound to particulate or suspended sediments
(mechanical weathering) can result in an increase of major ions in receiving waters (Schlesinger,
2-4

-------
1997) (see Figure 2-1). Most mines manage water and wastewater to minimize impacts on water
quality.
Climate change can also contribute to increased salinity of freshwater from increased
evaporation, intrusion through groundwater, and mobilization of geological salt deposits by
changes in aquifer charge and recharge with increased rainfall. Global climate change is often
linked to sea-level rises and intrusion of saltwater attributed to changes in pressure, expansion of
oceans as water temperatures increase, and glacial melting (Werner and Simmons, 2009).
Expansion and creation of estuarine tidal channels over time, from both anthropogenic and
natural causes, and compaction of plain lands have been found to contribute to saltwater
intrusion (Mulrennan and Woodroffe, 1998). Storm surges and flood tides in which water levels
exceed normal high tide levels may also contribute to saltwater intrusion (Zhichang et al., 2001).
Saltwater intrusion has been well documented in coastal areas of the United States
(Barlow and Reichard, 2010). Saltwater intrusion most commonly occurs as groundwater is
removed and seawater infiltrates aquifers, potentially contaminating drinking water supplies and
streams via groundwater discharge. Saltwater intrusion into freshwater systems can also be
attributed to or exacerbated by road construction projects and culverts (Stewart et al., 2002).
Waters used for irrigation mobilize salts within the soil and may increase the ionic
concentration of surface waters near agricultural fields. Agricultural irrigation return waters
contain a variety of salt ions based on the water source, natural chemical composition of the soil,
and ions associated with nutrient enrichment (NO3 , NH4+, and PO43 ). Ions including Na+, CP,
F , Mg2+, and SO42 have been shown to mobilize in soils in the western United States leading to
increased salinity of adjacent waterways and aquifers (El-Ashry et al., 1985; Leland et al., 2001;
Scanlon et al., 2009). These processes are influenced by changes in the amount and patterns of
rainfall and changes in climate. Elevated salinity is estimated to affect 10% of the world's
irrigated lands (Duncan et al., 2008) and may increase as climates become more arid.
Salts are commonly used during periods of snow and freezing weather as a method for
deicing roadways. The most common deicing agent is rock salt mainly in the form of sodium
chloride (NaCl), though other compounds are available, such as calcium chloride (CaCh),
magnesium chloride (MgCh), potassium acetate (KCH3CO2), or calcium magnesium acetate
(CaMg(CFbC02)4) (Novotny et al., 2008; Forman and Alexander, 1998). The use of rock salt on
snow and ice covered roads has increased salt usage in the United States from 163,000 metric
2-5

-------
tons in 1940 to more than 23,000,000 metric tons in 2005 (Novotny et al., 2008), primarily in the
northern states. As snow and ice melt, salt is transported via surface runoff to lakes and streams,
or groundwater via recharge and has been found to increase concentrations of ions in
surrounding waters (Blasius and Merritt, 2002; Novotny et al., 2008; Godwin et al., 2003).
Water quality impacts can be important because of the greater percentages of pavement in
urbanized watersheds. Salinity associated with deicing commonly occurs as seasonal pulses, as
materials are applied during freezing conditions and are transported into waterways upon
melting. However, in some areas, increased salinity attributed to deicing salts may persist in
surface waters due to delayed transport of salts stored in soil and groundwater from previous
winters (Jackson and Jobbagy, 2005, Kaushal et al., 2005).
Wastewater treatment plants and industrial discharges can contribute ions to freshwater
systems and can dominate water quality in streams and rivers dominated by effluent discharge.
Wastewater treatment plants have been shown to increase concentrations of Na+, CP, K+. Total
Kjehldahl nitrogen (TKN), SO42 , and SC downstream of the treatment plant discharge
(Andersen et al., 2004). Kaushal et al. (2005) found increasing concentrations of chloride in a
long-term study of streams. Echols et al. (2009) measured SC below a point source brine
discharge, which ranged from 5,900-18,000 [j,S/cm. Other industries including food processing,
petroleum, and leather production also produce saline wastewaters as a byproduct of production
(Lefebvre and Moletta, 2006).
Wright et al. (2011) have identified weathering of cement as a source of Ca2+ and HCO3
in streams draining urban areas. Rose (2007) also found these ions along with others to be
elevated in urban subbasins.
Some specific examples of anthropogenic sources of ions illustrated in Figure 2-1
(adapted and updated from Ziegler et al., 2010) and their associated dominant ions are
summarized in Table 2-1.
2-6

-------
Table 2-1. Examples of ions associated with different anthropogenic sources
Source
Dominant ions
References
Surface coal mining and valley fills
associated with mountaintop-removal
coal mining
Ca2+, Mg2+, HCO, .
SO,2
Bryant et al. (2002), Pond et al. (2008), EPA
(2011a, b), Griffith et al. (2012)
Runoff and effluents from conventional
coal mining and processing
Ca2+, Mg2+, HCO, .
SO.,2
Zielinski et al. (2001), Kennedy et al. (2003),
Kimmel and Argent (2010)
Deep coal mining
Na+, Ca2+, Mg2+, CI,
SOr
Thomas (2002), Mayhugh and Ziemkiewicz
(2005)
Combustion effluents
Ca2+, Mg2+, HCO, .
SO,2
Samarina (2007), Ruhl et al. (2012)
Historical industrial sources, such as
chlor-alkali plants
2
f+
0
1
Echols et al. (2009)
Wastewater treatment plants
Na+, CI, K+, TKN,
SO,2
Paul and Meyer (2001), Andersen et al. (2004)
Sewage and industrial waste discharges
Na+, cr, nh4+, no, .
PO43-
Carey and Migliaccio (2009)
Salt water intrusion
2
f+
O
1
Barlow and Reichard (2010), Mulrennan and
Woodroffe (1998), Barlow (2003)
Produced water from coalbed methane
production
Na+, HCO, . CI
Brinck et al. (2008), Dahm et al. (2011),
Jackson and Reddy (2007), National Research
Council (2010), Clark et al. (2001), Veil et al.
(2004)
Produced water from shale gas
production (i.e., hydrofracking)
Produced water from conventional
production of crude oil or natural gas
Na+, Ca2+, Mg2+, CI",
HC03 , K+, SO ,2. Br
Na+, CI
Haluszczak et al. (2013), Entrekin et al.
(2011), Gregory et al. (2011), Veil et al.
(2004)
Meyer et al. (1985), Boelter et al. (1992), Veil
et al. (2004)
Agricultural runoff, particularly
associated with irrigation
Na+, Mg+, NH4+, Cr,
F", SO ,2 . PO ,3
Ions may vary by
region.
El-Ashry et al. (1985), Leland et al. (2001),
Bernot et al. (2006), Lerotholi et al. (2004),
Lenat (1984)
Road deicing treatments
Na+, Cr, Ca2+, Mg+
Forman and Alexander (1998), Kelly et al.
(2008), Environment Canada and Health
Canada (2001), Evans and Frick (2001),
Kaushal et al. (2005)
Impervious surfaces and weathering of
concrete in urban drainage systems
Ca2+, hco3 , cr
Kelting et al. (2012), Steffy et al. (2004)
Wright et al. (2011), Rose (2007)
Dry and wet acid deposition
Ca2+, Mg2+, HCO, .
SO,2
Kaushal et al. (2013)
2-7

-------
2.2.2. Conceptual Model
A conceptual model consists of a written description and diagram that illustrates the
relationships between human activities, stressors, and ecological effects on assessment endpoints
(U.S. EPA, 1998a). The conceptual model links exposure characteristics with the ecological
endpoints important for management goals.
The simplified conceptual model shown here (see Figure 2-1) summarizes natural and
anthropogenic sources of ionic loadings in the case example study areas, transport pathways, and
potential ecological responses, all of which are described in greater detail in the following
sections. Sources are affected by processes or states that can result in delivery of a source to a
proximate stressor to the aquatic system. Sources deliver stressors, in this case, dissolved ions to
streams. The proximate stressor is the physical, chemical, or biological agent that directly causes
one or more biotic responses of concern, in this case, an increase in ionic concentration and/or a
change in the relative amounts of ions dissolved in the water. The physical biological exposure
is the form or route of exposure or uptake, which is generally direct contact with semipermeable
membranes such as gills and internal integument. The physiological mechanism is the
molecular, cellular, tissue, or organ system alteration that results from exposure to the stressor.
These include changes in ionic concentration, pH shifts, and possibly loss of epithelial integrity.
The mode of action is the organismal effect that may reduce fitness and survivorship and
increase emigration. The assessment endpoint is the adverse population level of effect, in this
case, extirpation. Extirpation is the depletion of a population of a species to the point that it is no
longer a viable resource or is unlikely to fulfill its function in the ecosystem (U.S. EPA, 2003).
The threshold for extirpation is operationally defined by the level below which 95% of the
observations of the genus occur, an XC95. For a more general model showing other sources,
such as marine intrusion associated with water withdrawal or fires resulting in ash, see the
conceptual model for ionic concentration on the CADDIS website
(http://www.epa.gov/caddis/ssr ion4d.htmD.
2-8

-------
c
geologic sources
)
dissolution of crushed rock
land cover alteration^)
atmospheric deposition J
f' waste water
Sources
'f' delivery of ions to stream
f ion concentration and A ionic composition
£
& ion
exchange

•f* competition for
anionic gill sites
3fL
^ osmotic
gradient
Delivery of stressor
Proximate stressor
Physical-biological exposures

A internal
pH
A ionic
gradients
A volume
control
4< essential
elements
toxicity of\ Physiological mechanisms,
specific ions/	Tissue |eve| effects
A behavior
4* growth
•f' migration
¦4, survivorship
A development
A metabolism
4' reproduction
Modes of action,
Organism level effects
decline in occurrence and
extirpation of macroinvertebrate genera _
Population level
effects
decline in abundance and
extirpation of some fish species
decline in other species
Figure 2-1. Conceptual model showing hypothesized relationships among
selected sources of ions and biotic responses to ionic stress by salt intolerant
taxa (adapted from Schofield and Ziegler, 2010).
Upward arrows indicate an increase, downward arrows indicate a decrease, and
delta symbols indicate a change in the parameter in either direction depending on
conditions. Inclusion of a linkage indicates that the linkage can occur, not that it
always occurs.
2.2.3. Environmental Transport and Fate of Ions in the Aquatic Environment
The majority of calcium (Ca2+) and magnesium (Mg2+) found in most soils and surface
water originates from chemical weathering of common minerals in rock or soils, such as
limestones (CaCCb) and dolomites (CaMg(C03)2) (e.g., Goddard et al., 2007). Minerals rich in
calcite, e.g., apatite (Cas(P04)3(F,Cl,0H)), can be found in igneous, sedimentary, and
2-9

-------
metamorphic rocks (e.g., Nezat et al., 2008). In many areas, these calcium and magnesium rich
rocks are relatively easily weathered and soluble, with their mobility strongly affected by pH,
becoming more mobile with decreasing pH (Likens et al., 1998). In forested catchments, the
calcium and magnesium concentrations in surface waters can increase following disturbances,
such as deforestation (Likens et al., 1970), and decrease in late successional forest stands relative
to early successional forest stands (Hamburg et al., 2003). In general, anions (negatively charged
ions) are more mobile than cations (positively charged ions) because they are not bound to
negative binding sites on clays. Bicarbonate (HCO3 ) ions in most soils and groundwater result
from chemical weathering of calcareous minerals. Bicarbonate ions are also present in soils as a
byproduct of plant and microbial respiration, as well as from the oxidation of organic matter
whereby carbon dioxide released in the soil becomes hydrated to form carbonic acid (H2CO3)
and is then dissociated into bicarbonate (HCO3 ) and carbonate (CO32 ), depending on the local
soil pH. The relative concentration of HCO3 compared to H2CO3 and CO32 is pH dependent,
with HCO3 being the dominant form at circumneutral pH. HCO3 is readily leached from soils
during rainfall. Alkalinity is a measure of HCO3 and CO32 .
Sulfate (SO42 ) ions found in soil and rocks can result from chemical weathering of
sulfate minerals, such as gypsum (Mullins and Hansen, 2006) or from chemical weathering of
coal deposits (Schlesinger, 1997). Atmospheric deposition can also be a source of sulfate found
in soils and is primarily anthropogenic in origin from the burning of fossil fuels (Schlesinger,
1997). Sulfate is readily leachable in soils, and sulfate mobility was found to be positively
correlated with rainfall in relatively undisturbed forested watersheds in both Central
Pennsylvania (Lynch and Corbett, 1989) and the Georgia Piedmont (Huntington et al., 1994). In
the Allegheny River Basin in southwestern Pennsylvania, sulfate concentrations in surface
waters draining relatively undisturbed watersheds ranged from 16-20 mg/L (Sams and Beer,
2000).
In addition to runoff, ions can be transported to surface waters through groundwater
discharge. Major ions can enter groundwater through dissolution of minerals in soils and rocks
during recharge. Particularly during periods of low streamflow, groundwater discharge can be a
major contributor of ions to surface waters (Larson and Marti, 1996).
Once mobilized, the majority of major ions that contribute to SC behave conservatively
in aquatic systems and are transported in surface water and groundwater to receiving waters.
2-10

-------
Although the ions that are the focus of these field-based methods are essential elements for living
organisms (within specific ranges), biological uptake does not effectively reduce ionic
concentrations in streams (U.S. EPA, 201 lb). In addition, Ca2+, Mg2+, SO42 , Na+, and K+ are
not significantly degraded nor adsorbed (U.S. EPA, 201 lb). As a consequence, concentrations
of the transported major ions tend to increase in receiving waters unless diuted by precipitation
or inflow from tributaries with lower ionic concentrations (Johnson et al., 2010; Merriam et al.,
2011). An exception is bicarbonate ions, which can be taken up by photosynthetic plants.
Geologically bound carbonates (HCO3 ) are also released into the atmosphere. Vesper et al.
(2016) reported a total flux of dissolved organic carbon from two sites near a coal mine that
ranged from 13 to 249 kg-C/year (18-364 metric tons of CCh/year).
2.3. MEASURE OF EXPOSURE
The concentration of a dissolved salt mixture can be measured in a number of ways: as
SC, total dissolved solids (TDS), freezing point depression (also referred to as osmotic pressure
or osmolality), refractive index, density, or the sum of the concentrations of individually
measured ions. A comparison of the capabilities of these different measurement methods is
shown in Table 2-2. The EPA has selected SC as the parameter to represent the measure of
exposure for this stressor. SC was selected as the measure of the ionic mixture for these
field-based methods because (1) SC is a measure of all ions in the mixture; (2) the measurement
technology is fast, inexpensive, accurate, and precise; (3) it measures only dissolved ions; (4) it
can be used to provide continuous monitoring records with in situ instrumentation; (5) it is a
sensitive measure in dilute waters; (6) it is less influenced by other nonfilterable material such as
oils and carbohydrates that may be dissolved in water compared to some measurement methods
(e.g., TDS); and (7) it is monitored by most state water monitoring programs at bioassessment
sampling sites. Several approved methods for measuring SC are available, including EPA
method 120.1 (U.S. EPA, 1982 revised).
SC has been commonly used as a measure of ionic concentration, and as an estimate of
major solute concentrations and total dissolved solids concentrations of natural waters
(McCleskey, 2011; Ziegler et al., 2010). SC is a measure of a material's ability to conduct
electric current, including natural waters, and is typically expressed in units of microsiemens per
centimeter ([j,S/cm).
2-11

-------
Table 2-2. Comparison of methods to measure ionic concentration
Measurement
method
All ions?
Speed
Approximate
sample range
and sample
volume
Sample
filtration
required
Field
use
Continuous
measure
possible
Affected by
nonionic
constituents
Specific
conductivity
Yes
Seconds
Wide range,
pL-mL,
volume or
in situ
No
Yes
Yes
No
Total dissolved
solids
(gravimetric)
Yes
Days
Requires larger
volumes for
freshwater
At times
No
No
Yes
Freezing point
depression
Yes
Minutes
Wide range,
few pL to mL
volumes
At times
No
No
Yes
Refractometry
Yes
Minutes
Better suited
for higher
salinities, pL
volumes
At times
Yes
Industrial
application
Yes
Densitometry
Yes
Minutes
Better suited
for higher
salinities, dl
volumes
No
Yes
No
Yes
Sum of ion
concentrations
Typically
major ions
only; e.g.,
Ca2+, Mg2+,
Na+, K+,
cr, so42,
and HCO3
Hours to
days
Variable
depending on
analytical
methods
Yes
No
No
No
Because SC predictably increases with increasing ionic concentration, it is used to
measure salinity (usually referring to NaCl) or ionic concentration (for any dissolved salts)
(Standard Methods #2510 [APHA, 1992]; EPA method 120.1, 0950A [U.S. EPA, 1982]). SC
measurements in natural waters indicate the presence of inorganic dissolved solids
(e.g., chloride, nitrate, sulfate, bicarbonate, nitrite/nitrate, and phosphate anions and sodium,
potassium, magnesium, calcium and iron cations). Electrical currents are carried by both
positively charged cations and negatively charged anions—but to differing degrees depending on
charge and mobility. Thus, the SC of a mixture depends on the type and concentration of the
2-12

-------
ions in solution. SC is also dependent on temperature and is known to increase approximately
2% for every 1°C increase in water temperature. The term "Specific Conductivity" indicates the
measurement has been standardized to 25°C, a reference temperature (Wetzel, 2001). SC is
commonly reported in state monitoring programs, rather than the unstandardized conductivity
measurement.
Both specific conductivity and specific conductance are often used synonymously in the
open literature indicating normalization or measurement at 25°C. Conductivity is a property of
water expressed as [j,S/cm. Conductance of a sample or electrical component is measured as
Siemens (S). All measurements in this document refer to specific conductivity/specific
conductance expressed as [j,S/cm at 25°C as it relates to water samples.
SC is an aggregate measurement of the full ionic mixture of a water sample. The total
ionic concentration of natural waters is associated with biological effects. However, waters with
similar SC levels may have different ionic compositions, and as a result can have different
toxicities to freshwater organisms in the laboratory and in the field (Mount et al., 2016; Zalizniak
et al., 2006; Dunlop et al., 2015). Therefore, when using SC as a measure of ionic concentration,
it is important to characterize the specific ions and their relative concentrations.
Some states and authorized tribes may want to use an alternative measurement of ionic
concentration when developing aquatic life criteria. If a different measure of the ionic mixture is
selected as the measure of exposure, the reliability of the measurement should be considered.
For example, TDS has greater variability than other methods. If some states and tribes prefer to
measure specific ions known to be toxic to aquatic organisms, the interaction of ions within the
mixture also needs to be considered. Appendix F provides an example using an alternative
measure of exposure for waters dominated by Ca2+ and Mg2+, the sum of HCO3" and SO42 in
mg/L.
2.4. NATURE OF I II I EFFECT
All tolerances of stressors are determined by the evolutionary adaptations of organisms.
The background levels of naturally occurring habitat variables such as temperature, pH, and SC
are important determinants of those adaptations. Because aquatic species evolved in unpolluted
waters, background levels define aspects of the niche to which the biota of a community is
naturally adapted and which it potentially tolerates (MacArthur and Levins, 1967; Colwell and
2-13

-------
Rangel, 2009; Peterson et al., 2011; Futuyma and Moreno, 1988; Wiens, 2004). Aquatic species
inhabit nearly pure water, estuarine and marine conditions, hypersaline pools, and everything in
between (Remane, 1971, Potapova and Charles, 2003; Potapova, 2005; Berra, 2007). In most of
the United States, freshwater habitats have very low concentrations of dissolved ions relative to
marine systems, so that is the condition to which most freshwater biota are adapted.
Algae, protozoans, zooplankton, and bacteria have all been shown to have SC preferences
in freshwater systems (Potapova and Charles, 2003; Potapova, 2014; Bos et al., 1996).
Nationally, of 230 soft-bodied algae identified to the lowest practical taxonomic level, 56% had
estimated optima <500 [j,S/cm (Potapova, 2014). Nationally, of 683 diatoms also identified to
the lowest practical taxonomic level, 84% had optima <500 [j,S/cm and 35% did not occur in
water >500 [j,S/cm (Potapova, 2014).
Freshwater benthic macroinvertebrates are extirpated at different ionic concentrations
(U.S. EPA, 201 la). In West Virginia, 17% of genera that occur at background SC are extirpated
at 500 [j,S/cm and many more genera decline at that SC (Cormier et al., 2013b). Effects are not
limited to Appalachia. In Nevada streams, differences between observed and expected
invertebrate communities increased above natural background levels of approximately
300 [j,S/cm (Vander Laan et al., 2013). Freshwater fish also decline and are extirpated as ionic
concentration increases (see Appendix G). Although these data are from waterbodies with a
wide range of background SC values, they demonstrate that many species and genera are adapted
to particular SC regimes, and many of them are quite low.
The physiological limits of species determine their tolerance ranges, in this case, their
potential SC niche with respect to concentrations of a defined ionic mixture (Olson, 2012;
Vander Laan et al., 2013). At the extremes of their physiological tolerance, species are less able
to develop, grow, and reproduce. A species may not exploit its full tolerance range, because
competitor species are better suited for a particular ionic concentration or for other ecological
reasons such as predation, parasitism, and habitat requirements. The SC range that is actually
inhabited by a species is called a realized niche.
The range of SC conditions varies in natural aquatic systems. Species do not occur where
the SC is lower or higher than their SC tolerance. The lowest SC in a freshwater system,
therefore, is the lowest possible limit of the potential SC niche (see lower tolerance limit in
Figure 2-2). When mineral salts are added to an aquatic system, SC increases, part of the
2-14

-------
potentially habitable SC niche space is lost, and the size of the realized niche for species adapted
to low SC decreases. When the SC is above the physiological tolerance of a species due to
natural or anthropogenic causes, it does not persist and the species is extirpated.
Ephemerella
XC95
CuO <9
sink
habitat
realized00
niche
D.
100
1000	10000
Specific Conductivity (|_iS/cm)
lower
tolerance
limit
Figure 2-2. A species' (or genus') realized niche is defined by its lower and
upper limits of occurrence. In this case, the lower tolerance limit is less than or
equal to the lowest specific conductivity (SC), which is the lower limit of
occurrence. The XC95 represents the upper tolerance limit. Approximately 5% of
observations of a taxon are assumed to occur in sink habitats where a population
cannot persist without immigration from source habitats. A species or genus
optimum is the environmental condition most easily tolerated both
physiologically and competitively and can be estimated by the conditions where
the taxon is most often observed. The optimum SC may be estimated at the
maximum probability of observing the taxon from a generalized additive model,
shown here to be the minimum SC. The example involves the genus Ephemerella
which is comprised of several species of mayflies.
The upper tolerance limit of a species is estimated by its XC95 (see Figure 2-2).
Extirpation is the depletion of a population of a species to the point that it is no longer a viable
resource or is unlikely to fulfill its function in the ecosystem (U.S. EPA, 2003). The occurrences
of benthic invertebrate species at locations with a SC greater than their XC95 value are believed
2-15

-------
to represent sink habitats (Pond et al., 2014). Sink habitats are those locations where occurrence
of species is primarily the result of immigration from locations with low SC termed source
habitats from which immigrants originate. They are "sinks" in the sense that immigrants have
low success in establishing sustainable populations in those locations.
These phenomena have practical application. The proportion of species or genera
extirpated as a result of increased SC in an ecoregion can be determined and is the basis for the
XCD method.
Several other predictions can be made from niche theory. Species with niches that limit
them to low SC water are not expected to occur where low SC water does not occur. The source
of high SC could be natural or due to anthropogenic inputs (Cormier et al., 2012, Coffey et al.,
2014). For example, in an ecoregion lacking streams <400 [j.S/cm, any species with an upper
tolerance limit <400 [j.S/cm SC would not be expected to occur because there is no habitat for
them. As a corollary, where there is a low SC habitat in an ecoregion, species tolerant to low SC
will occur.
The relationship between ambient SC levels and SC tolerances of species that are present
has at least two practical implications. First, it is inappropriate to set criteria below natural
background for a location. Second, the lower limit for any XCD in any given ecoregion cannot
be lower than the natural background of the ecoregion. In practical terms, this shifts the origins
of XCDs and their 5th centiles toward higher SC (graphically to the right) as the background SC
increases. Hence, when XCDs from regions of low to high natural background are
simultaneously plotted on the same graph, the curves progress to the right. (For an example, see
the XCDs in Appendix D, Figure D-3). Therefore, the background SC of an ecoregion is
strongly associated with a predictable extirpation of 5% of species or genera. This relationship
between background SC and the proportion of extirpation can be used to predict the SC that will
extirpate 5% of species or genera in an ecoregion solely based on ecoregional background (see
Section 3.7.2 and Appendix D, Figure D-4).
2.5. MECHANISMS AND MODES OF ACTION
The measure of effect for these field-based methods is extirpation (U.S. EPA, 201 la). The
three most likely modes of action for extirpation of a genus or species are the population-level
processes mortality, emigration, and failure to recruit (Rubach et al., 2011; Williams and Hynes,
2-16

-------
1976; Clements and Kotalik 2016). The sections below discuss some physiological mechanisms
of action through which SC acts on organisms and on the processes that constitute the potential
modes of action.
2.5.1. Physiological Mechanisms
In exposures to elevated ionic concentrations, physiological stress could cause mortality
or drift (a process in which invertebrates emigrate by releasing the substrate and allowing
themselves to be carried downstream). The stress occurs because the freshwater organisms
cannot maintain or need to use more energy to maintain their internal ionic concentration and pH
with altered ionic composition and concentration, and water volume in waters with very high
ionic concentration. The mechanism of action is believed to be due to adverse ionic gradients
formed by the concentration and relative proportions of ions. For all freshwater organisms,
microbes, plants and animals alike, ionic concentration is higher inside an organism than in
freshwater. To concentrate and maintain the internal ion concentration, organisms have evolved
many interrelated strategies. One cannot describe the specific action of toxicity of one ion or pH
without considering all the others (Zhang and Wakamatsu, 2002 Griffith, 2016; Bradley, 2009;
Evans, 2008a, b; Wood and Shuttleworth, 2008; Nelson and Cox, 2005; Marshall, 2002; Hille,
2001; Smith, 2001; Thorp and Covich, 2001; Komnick, 1977; Sutcliffe, 1962). For example,
Na+ and CP concentrations are much higher inside organisms than in freshwater. One
mechanism used by invertebrates and fish to concentrate CP, an anion with a negative charge, is
to exchange CP for a nonmineral anion waste product (CO2) that is produced during metabolism
of sugar. An enzyme, carbonic anhydrase, rapidly and reversibly catalyzes water and C02to
HCO3 and H+. HCO3 concentrations are higher inside the organism and lower in the water.
This concentration gradient is favorable for the exchange of CP. However, a cation also needs to
be removed from the organism or else H+ will accumulate and cause acidosis. Acidosis causes
complex cellular reactions and affects function of cellular organelles that lead to many adverse
effects including death (Gesser and Poupa, 1983; Vafai and Mootha, 2012). Freshwater animals
exploit this increased concentration of H+ by exchanging it for another cation, such as Na+.
Thus, Na+ and CP are concentrated inside organisms relative to freshwater. However, when the
HCO3 concentration in freshwater is high, the concentration gradient does not favor movement
ofHC03 out of the organism and other ions are not readily brought into the organism. Because
2-17

-------
this anion-cation exchange mechanism uses waste CO2, it requires less energy to maintain.
Low-energy regulation of ions that depend on favorable HCO3 concentrations can be
supplemented by adenosine triphosphate-dependent transport of ions as ion concentrations
increase outside the organism and concentration gradients become less favorable for passive or
low energy transport. The inability to regulate internal ionic concentrations or the greater energy
demand for ion regulation may causes stress resulting in death, drift, reduced growth, or reduced
reproduction, but definitive cellular studies for most aquatic organisms are lacking.
Organisms use many strategies to minimize loss of ions and the exclusion of water (see
references in previous paragraph). At the interface between water and the organism's surface,
epithelial tissue integrity is essential. Cell membranes are a barrier to water because they are
hydrophobic bilayers of lipids. The membranes are selective for the ions and direction of
movement using proteinaceous ion channels, ports, and carriers (for a review see Griffith, 2016).
Between the cells making up the epithelial pavement, ultrastructural features called tight
junctions hold adjacent cells together and complete the epithelial barrier restricting water and ion
movement into or out of the organism. External Ca2+ helps maintain tight junctions (Gonzales
and McDonald, 1992; Smith et al., 2005; Brown and Davis, 2002). There is some evidence from
human studies of the gut that SO42 may interfere with tight junctions causing loss of epithelial
integrity but the physiological interactions of SO42 have not been well studied in freshwater
organisms. Note that ion concentrations in freshwater are always less than inside the animal and
do not cause loss of water from the animal. Rather, loss of epithelial integrity can lead to excess
water or loss of ions. This is a key difference between marine and freshwater organisms.
In summary, the full complement of anions and cations, including others not described
here, need to be maintained by organisms. There is an extensive literature on ionoregulation of
cations and anions. The higher concentration of ions inside organisms compared to freshwater
provides opportunities to use ionic gradients for ionoregulation. Acid-base regulation is linked
to the production of hydrogen ions involved in ionoregulation. Because useful gradients are
dependent on low concentrations of ions in freshwater, relative amounts of each ion, not
necessarily any individual ion, accounts for toxic effect.
2-18

-------
2.5.2.	Mortality, Growth, and Reproduction
Death of juvenile aquatic invertebrates exposed to different ionic concentrations has been
demonstrated in the laboratory (Echols et al., 2010; Kennedy et al., 2003, 2004, 2005; Lasier and
Hardin, 2010; Merricks et al., 2007; Mount et al., 1997; Wang et al., 2013, Kunz et al., 2013,
Bringolf et al., 2007). Sublethal effects reported from laboratory studies include reduced growth,
reproduction (Johnson et al., 2015), early emergence (Nietch et al., 2014), and premature release
of unionid glochidia (Gillis, 2011). When death of an entire population occurs, the area remains
depopulated until recolonized by aerial dispersion and egg-laying (oviposition) (Smith et al.,
2009) or by organisms floating downstream (drift) from refugia at upstream reaches or tributaries
to the depopulated stream reach (Williams and Hynes, 1976; Pond et al., 2014).
2.5.3.	Emigration
Emigration occurs when organisms vacate a stressful environment after being challenged
with a noxious stimulus or lack of food or other resources. In numerous studies, benthic
invertebrate drift is induced within minutes of exposures to a range of stressors in natural and
artificial streams (Svendsen et al., 2004; Wood and Dykes, 2002). Stress induced drift and
avoidance behaviors have been shown to occur with salts, toxic chemical spills, floods,
pesticides, drought, sediment, low dissolved oxygen (DO), heat, and organic pollution (Wood
and Dykes, 2002; Svendsen et al., 2004; Crossland et al., 1991; Doeg and Millage, 1991;
Wallace, 1990; Brittain and Eikeland, 1988; Sheehan and Winner, 1984; Geckler et al., 1976;
Waters, 1966, 1972, 1995). In independent studies, colonized substrates were exposed to
continuous flowing treatments of ionic mixtures (Clements et al., 2014, 2016; Nietch et al.,
2014). The studies showed increased drift, reduced numbers of taxa, and other effects. Drift is
more likely to occur when there is an abrupt change in environmental conditions rather than a
slow change that allows organisms to physiologically adapt. For example, after a moderate
increase in ionic concentration, some aquatic insects synthesize more ionic channels for ionic
regulation (Wichard et al., 1973; Sutcliffe, 1974; Komnick, 1977).
2.5.4.	Failure to Recr'uit
Development begins with gamete production. Fertilization during the terrestrial phase of
the life cycle occurs internally in most aquatic insects and is unlikely to be affected by aqueous
2-19

-------
ions. At oviposition, contact with freshwater causes the swelling and formation of an
extrachorionic coating that is necessary for adherence of the eggs of some invertebrates to
substrates in the stream (Percivale and Whitehead, 1928). Eggs oviposited into water with
higher specific SC do not form the adhesive coating (Percivale and Whitehead, 1928) and the
eggs are washed downstream and presumably perish (Gaino and Bongiovanni, 1992). Ionic
gradients that initiate biological changes are also necessary to permit propagation of a
fertilization potential over the surface of some eggs and to allow successful embryonic
development and hatching in some species such as fish (Jaffe, 1991; Coward et al., 2002).
Similarly, toxicity tests with fathead minnow larvae were more sensitive during the transitional
period from embryo development to hatching (Wang et al., 2016a). Mesocosm experiments with
mayflies also indicated greater vulnerability during early life stages and during emergence to
winged adults (Clements and Kotalick, 2016; Nietch et al., 2014).
2.5.5. Community Interactions
Increased competition, predatation, or parasitism have been suggested as possible modes
of action leading to loss of some species and an increase in salt-tolerant taxa where ions are
elevated (Olson, 2012; Olson and Hawkins, 2012; Micieli et al., 2012; Wood-Eggenschwiler and
Barlocher, 1983). These processes may affect the benthic invertebrate communities that form at
different ionic concentrations.
2.6. ASSESSMENT ENDPOINTS AND MEASURES OF EFFECT
2.6.1. Assessment Endpoints
Assessment endpoints represent the actual environmental value to be protected. They are
defined by an ecological entity (e.g., species, community, or other entity) and attributes,
(e.g., survival, growth, and reproduction) (U.S. EPA, 1998a). In the development of water
quality criteria for SC, the entities are aquatic biotic communities and the attribute is protection
of all but a small fraction of species from extirpation.
The relevant ecological entities for these field-based methods are macroinvertebrate
assemblages, which are characterized by their taxonomic composition at the genus level.
Macroinvertebrates were selected because they are susceptible to ionic stress, they are important
to stream function and ecosystem integrity, they provide numerous ecosystem services that
2-20

-------
benefit humans, they can be found in all types of streams, and they are intrinsically valuable
aquatic life forms (Suter and Cormier, 2015). Furthermore, because macroinvertebrates
constitute the great majority of multicellular species in streams and have a wide range of
sensitivities, they are excellent indicators of adverse effects on ecological processes and on the
larger aquatic community. For these reasons, all states and many tribes monitor aquatic
macroinvertebrates to assess the health of the aquatic community (U.S. EPA, 2002).
The most commonly recognized contribution of aquatic macroinvertebrates is that they
are food for larger invertebrates and fish and other vertebrates, including recreationally important
fish species (Allan, 1981; Richardson, 1993; Sweka and Harman, 2008; Hitt and Chambers,
2014), amphibians (Burton, 1976; Wallace et al., 1997), insectivorous bird species (Nakano and
Murakami, 2001; Gray, 1993; Epanchin et al., 2010), bats (Clare et al., 2011), and mammals.
However, the overall function of freshwater aquatic ecosystems is also dependent on
macroinvertebrates (Hooper et al., 2005; Cardinale, 2011). Macroinvertebrates improve water
quality through forest and stream nutrient retention (Newbold et al., 1983, 1982; Wallace and
Webster, 1996; Huryn and Wallace, 2000; Evans-White et al., 2005), aid in leaf litter
decomposition (Wallace and Webster, 1996), and remove pathogens and nuisance periphyton
blooms by filtering and grazing (Wallace and Merrit, 1980; Yasuno et al., 1982; Hall et al.,
1996). Because macroinvertebrates provide many ecosystem services, it is well understood that
stream macroinvertebrate diversity and abundance are important indicators of overall stream
condition (Carter et al., 2006; Resh, 1995), and many stream monitoring programs and stream
condition indices rely on macroinvertebrate sampling metrics (Gerritsen et al., 2000; Pond
etal., 2008; U.S. EPA, 2002).
These field-based methods can be used to develop ecoregional criteria that are fully
protective of aquatic life. Many freshwater insects are among the most salt-intolerant organisms
relative to other taxa, including crustaceans such as crayfish and daphnids, fish, and amphibians
(compare Appendices A.4 and B.4 with Appendix G of this report). Recent studies suggest that
mussels in the family Unionidae are acutely salt-intolerant (Kunz et al., 2013; Wang et al.,
2016a, b), particularly during early (glochidia and juvenile) life stages (Bringolf et al., 2007;
Gillis, 2011; Wang et al., 2016a, b).
Fish also are adversely affected by ionic stress (see Appendix G, Stauffer and Ferreri,
2002; Kimmel and Argent, 2010; Mount et al., 1997; Kennedy et al., 2005, 2004, 2003; Harper
2-21

-------
et al., 2012; Farag and Harper, 2012, Hopkins and Rousch, 2013, Hitt and Chambers, 2014).
EPA's assessment of fish in Appendix G indicates that they are sensitive to ionic stress but are
extirpated at slightly higher SC levels than macroinvertebrates. Therefore, fish are expected to
be protected by criteria based on macroinvertebrate data. More complex organisms (e.g., fish)
generally have a greater ability to regulate internal ionic concentrations and water volumes than
simpler organisms, such as benthic invertebrates (Dunlop et al., 2005). Fish also have greater
mobility and may be able to more readily migrate from high SC sites to more habitable areas
(e.g., Goldstein et al., 1999; Woodward et al., 1997, 1995).
In sum, macroinvertebrates are a critical component of ecological integrity, provide
numerous ecosystem services, and appear to be a salt-intolerant ecological taxonomic group;
therefore, they are used as an assessment endpoint for these field-based methods.
2.6.2. Measures of Effect
The measures of effect for these field-based methods have been selected to be consistent
with the intent of the conventional laboratory-based method for developing aquatic life criteria
(U.S. EPA, 1985). Two relationships are derived from the paired SC and macroinvertebrate field
data: one for each macroinvertebrate genus and one for the overall macroinvertebrate community
in the study area. First, a cumulative distribution function (CDF) is developed for each genus2 to
determine its genus XC95, the SC level above which a genus is effectively absent from water
bodies in a region (U.S. EPA, 2003). It is defined in this method as the 95th centile of the
distribution of occurrences of a macroinvertebrate genus. In other words, the probability of
observing a genus above its XC95 SC value is 0.05; i.e., if a genus is observed at 100 sites, only
5 sites would be expected to have SC above the XC95. XC95 values that are uncertain or
unmeasured within the exposure range are noted and generally do not influence the hazardous
concentration (HC05) because their estimated XC95 values are greater than those genera in the
5th centile. Second, the HC05 is developed using a genus-level XCD for the macroinvertebrate
community from the aggregation of the XC95 values.
9
Conventionally, species have been aggregated to the genus level. However, effect levels may be different for
species within a genus due to niche partitioning afforded by naturally occurring causal agents such as dissolved ions.
(Remane, 1971; Suter, 2007). Hence, an apparently salt-tolerant genus may contain both salt-intolerant species and
tolerant species. Analyses with fish species indicate that the range of XC95 values within a genus can be quite broad
and the empirical genus-level XC95 tends to represent the maximum XC95 of the species in the data set (see
Appendix G).
2-22

-------
One key difference from laboratory-based methods used to develop aquatic life criteria
for chronic exposures is the measure of effect. In the XCD method, the measure of effect is
genus extirpation (population-level) rather than an effect at the organism level. In the
laboratory-based method, the measures of effect represent survival, growth, or reproduction
(U.S. EPA, 1985). Because the example ecoregional criteria are based on field data for a large
number of macroinvertebrate genera across many sites across a broad SC gradient, the EPA
anticipates that a reasonable level of protection of the overall aquatic community will be
provided if all except a small fraction (i.e., 5%) of sampled macroinvertebrate genera from the
region are protected. In their review of the EPA Benchmark Report, the EPA SAB stated that
this approach provides a degree of protection comparable to or more protective than a
conventional water quality criteria based on conventional chronic toxicity testing
(U.S. EPA, 2011c).
The genus-level XCD used in the XCD method represents the response of genera in
biotic communities in general to a stressor (e.g., an ionic mixture dominated by sulfate plus
bicarbonate). XCDs do not require that the species or genera be the same in all applications or at
all locations (Posthuma et al., 2001; Cormier and Suter, 2013a; Cormier et al., 2013a). Similarly,
the genera that form the minimum data set for laboratory-based aquatic life criteria are not
intended to match any particular community; rather, they are surrogate taxa that represent any
potentially exposed freshwater community (U.S. EPA, 1985). In the same way, the distribution
of genera in the XCDs used in the XCD method (e.g., see Section 4.2) represent all stream
communities from a similar background ionic concentration exposed to a similar ionic mixture.
All of the macroinvertebrate taxa used to develop an XCD may not occur at any one site in an
ecoregion.
Because this approach relies directly on paired observations of in situ measurements of
SC and benthic invertebrate assemblage information, the potential adverse effects of ionic stress
on all life stages is considered in the context of other complex relationships (e.g., food web
dynamics) and aquatic ecosystem processes. The measures of effect (i.e., XC95 and HC05; see
Sections 3.1.2 and 3.1.3) are considered chronic-duration endpoints because the field data reflect
exposures over whole life cycles and multiple generations of the resident biota (see Table 2-3).
A field-based method to directly develop acute criteria for SC is not yet available due to a lack of
field data with sufficiently high temporal resolution (e.g., daily measurements of SC paired with
2-23

-------
macroinvertebrate sampling for at least 1 year). However, these field-based methods include a
method that uses within-site variability of SC levels to derive a criterion maximum exposure
concentration (CMEC) that will protect aquatic life from acutely toxic exposures (see
Section 3.2). The CMEC differs from a criterion maximum concentration (U.S. EPA, 1985)
because it is not calculated using laboratory or field data showing a direct relationship between
SC and an acutely toxic biological response. However, the protectiveness of the CMEC was
corroborated with field biological and SC data in Case Examples I and II (see Appendices A
and B).
Table 2-3. Summary of assessment endpoints and measures of effect used in
this field-based method to derive a criterion continuous concentration (CCC)
and criterion maximum exposure concentration (CMEC) for specific
conductivity
Stressor of concern
Measure of exposure
Mixture of ions (e.g., ([HCO3 ] + [SO42 ]) > | CI |
Specific conductivity
Assessment endpoints for the aquatic community
Measures of effect
Occurrence of macroinvertebrate populations
Chronic XC95 (genus-level effect)
Chronic HC05 (assemblage-level effect)
XC95 = Extirpation concentration, the SC value below which 95% of the observations of a genus occur.
To summarize, for these field-based methods, the valued resource is the aquatic
community, characterized by the macroinvertebrate populations that occur at a site. The
ecological entities defining the assessment endpoints are populations of macroinvertebrates
(aggregated to genera) and the measure of effect is extirpation (the desired attribute is
occurrence). Macroinvertebrate populations are appropriate assessment entities because they
occur in all but the poorest-quality streams, they are important to ecosystem structure and
function, they are highly diverse, they are common forms of aquatic life, and they are affected by
many different agents including ionic stress. Extirpation is the depletion of a population of a
species or genus to the point that it is no longer a viable resource or is unlikely to fulfill its
function in an assessed ecosystem. Extirpation of genera is an appropriate attribute for these
methods that rely upon analyses of field data to determine population-level effect thresholds.
Aquatic life criteria developed using these field-based methods are set at a SC level that protects
2-24

-------
95% of resident macroinvertebrate genera that occur at reference sites in the relevant ecoregion.
Aquatic communities are expected to be resilient to these effects and support the overall integrity
and function of the aquatic ecosystem if the derived ecoregional criteria (magnitude, duration
and frequency) are not exceeded.
2.7. SELECTION OF A FIELD-BASED METHOD
The EPA typically relies on laboratory toxicity test data for the development of aquatic
life criteria (U.S. EPA, 1985). To the extent laboratory toxicity tests have been performed on
similarly mixed proportions of these major ions, these tests with commonly tested laboratory
species have not indicated sensitivity at the concentrations associated with loss of
macroinvertebrate genera in the field (U.S. EPA, 201 la). Although it is impractical to replicate
in the laboratory the range of taxa, conditions, effects, or interactions that occur in natural
streams, a number of recent toxicity studies have begun to bridge the gap between physiological
studies and field observations (Mount et al., 2016; Erickson et al., 2016; Wang et al., 2016a;
Kunz et al. 2013). Also, mesoscosm studies have begun to identify the modes of action that
likely lead to extirpation in the field and corroborate effects observed in the field (Clements and
Kotalik, 2016; Nietch et al., 2014).
Analyses of field data show the reduced presence of many benthic macroinvertebrate
genera at increasing SC levels. The associations between SC and benthic macroinvertebrate
occurrence observed in the field have been assessed as causal (e.g., see Section 2.2 of this
document; [e.g., Gerritsen et al., 2010; Palmer et al., 2010; Lindberg et al., 2011; Merriam et al.,
2011; Pond et al., 2010, 2008; U.S. EPA, 2011a, b; Bernhardt et al., 2012; Cormier et al.,
2013b, c; Timpano et al., 2014; Dunlop et al., 2015; Zhao et al., 2016]). Furthermore, the field
data used in this analysis represent exposures to regionally representative assemblages of taxa
and life stages at levels and proportions of ions under realistic physical and chemical conditions.
In this case, field data can be used to directly assess ecologically-relevant measures of effects,
such as extirpation of genera in aquatic communities as a result of exposure to these ionic
mixtures in the field. Protection of a diverse assemblage of benthic macroinvertebrates is
protective of stream communities and aquatic life.
For these reasons, the EPA concluded that a field-based approach is appropriate for
developing SC criteria where there are sufficient data for analysis.
2-25

-------
3. ANALYSIS PLAN: FIELD-BASED METHODS TO DEVELOP SPECIFIC
CONDUCTIVITY CRITERIA
This section presents field-based methods that may be used to derive SC water quality
criteria for the protection of aquatic life where sufficient data are available. The methods
describe how to derive a criterion continuous concentration (CCC) to protect against effects from
chronic exposure (see Section 3.1) and a CMEC to protect against effects from acute exposure
(see Section 3.2). The methods also describe how to estimate duration (see Section 3.3) and
frequency (see Section 3.4), how to assess causation (see Section 3.5), and how to determine the
applicable geographic range of criteria (see Section 3.6). Because the primary method (the XCD
method) requires large, paired chemical and biological data sets which are not available for all
ecoregions in the United States, EPA developed another method which uses a model to calculate
a CCC for SC. This method is useful for Level III ecoregions where sufficient paired biological
data are lacking (see Section 3.7). Section 3 serves as the Analysis Plan for the case studies in
Sections 4, 5, 6 and 7. Some ecoregion-specific examples are included in this section to
illustrate key concepts and methods.
3.1. DERIVING A CRITERION CONTINUOUS CONCENTRATION (CCC)
The XCD method requires a field data set with paired in situ measurements of SC and
benthic macroinvertebrate survey results for streams in the study area. The field SC
measurements allow for the development of exposure-response relationships across a SC
gradient. The inclusion of high quality and impaired sites in the data sets assures a range of SC
(exposure) for characterizing changes in taxa occurrence (response). In aggregate, the in situ
field measurements of SC from many sites from different times of the year represent the
variability over a year.
Using this method, two relationships are derived from the field data: one for each
macroinvertebrate genus and one for the overall macroinvertebrate assemblage in the study area.
First, a weighted CDF is developed for each macroinvertebrate genus to determine each XC95.
Second, the HC05 is developed using a genus XCD for the macroinvertebrate community from
the aggregation of the XC95 values (an example is provided in Figure 3-1).
3-1

-------
"O
cu
¦M
ro
Ql
oo
o
ro cd
o
a»
c
a;
(D
c
o
¦"T
o
o
Q.
O
CNI
o
o
o
305 uS/cm
200
i	r
500	1000	2000
Specific Conductivity (|aS/cm)
Figure 3-1. Example of a genus extirpation concentration distribution
(XCD) depicting the proportion of genera extirpated with increasing
ionic concentration measured as specific conductivity (SC). Each point
on the XCD plot represents an extirpation concentration (XC95) value of
one genus arranged from the least to the most salt-tolerant. XC95 values
that were defined as greater than values are indicated by triangles. The
5th centile of the XCD is shown as a dotted horizontal line. The 5th centile
hazardous concentration (HC05) is the SC at that intercept of the XCD and
the 5th centile line. In this example, the HC05 is 305 [j,S/cm.
Several methods were evaluated prior to selection of a weighted CDF model to estimate
an XC. They included models of XC from logistic regression, a generalized additive model
(GAM), unweighted CDF, and other options. A weighted CDF model was selected because the
HC05 value fell within the range of the other methods, and it was computationally simple.
Weighting normalizes the distribution of samples taken across the SC gradient. The weighted
CDF does not assume any particular shape to the distribution and does not fit a function to the
3-2

-------
data points. Outliers were not identified or removed because little was known about the
sustainability of populations in high SC water or the movement of salt-intolerant taxa from
biological sources to sinks. However, at least one study indicates that apparent outliers may
represent transient occurrences of genera drifting from low SC to a high SC stream reach (Pond
et al., 2014). Removal of the outliers or using the area under a fitted curve such as a GAM
generally yields lower XC95 values and a lower HC05. As new information arises, the method for
modeling the XC95 may be updated. The EPA SAB endorsed EPA's selection (U.S. EPA,
201 lc) of the weighted CDF model for this field-based method (U.S. EPA, 201 la). The
statistical packaged, Version 2.12.1 (December 2010), was used for all statistical analyses in the
Case Studies (^-Development Core Team, 2011). The program "R" is open-source and
open-access computational software that runs on Microsoft Windows, Apple MacOS, and UNIX
platforms. The calculations can also be performed with a hand-held calculator or with a
spreadsheet such as Microsoft Excel.
Different forms of the exposure-response relationships (i.e., decreasing, unimodal,
increasing, and no relationship) are expected given the nature of the ions and the physiology of a
macroinvertebrate genus. For example, many ions are required for survival and are beneficial at
low levels but elicit toxic effects at high levels; such a stressor-response relationship is expected
to have a unimodal distribution (see Appendix A.3, Isonychia). In the ascending (left) limb,
requirements for ions are increasingly being met; in the descending (right) limb, toxicity is
increasing. However, many empirical exposure-response relationships for ions do not display
both limbs of the distribution. For example, some may show: (a) only the descending portion of
the curve because none of the observed SC levels are sufficiently low to show elemental
deficiency for the tax on (see Appendix A, Ephemerella and Leuctra); (b) only the ascending
portion because none of the observed SC levels are sufficiently high to show toxicity for the
taxon (see Appendix A, Cheumatopsyche)\ or (c) no trend at all because the optimum is more of
a plateau than a peak so it extends across the range of observed SC levels (see Appendix A.3,
Br ilia).
The steps involved in selecting, characterizing, and analyzing macroinvertebrate
sampling field data to derive a CCC are depicted in Figure 3-2 and described in the following
sections.
3-3

-------
Remove sites where pH <6 and sites that
have a different ionic composition
1/1
include reference genera that occur at >24 sites
Develop exposure-response models for each genus
Identify XC9S from CDF at genus extirpation
level (95th centiie of CDF)
Develop exposure-response model for all genera,
an extirpation concentration distribution (XCD)
Identify the biological effect level from
XCD at 5th centiie (HCos)
u
Round to 2 significant figures
XCD
field Data Set
Set of genus- CDFs
Paired data set: SC
and genus presence
Paired data set with
>90 genera
Criterion Continuous
Concentration
Figure 3-2. Main steps in the derivation of a chronic specific conductivity
(SC) criterion using the extirpation concentration distribution (XCD)
method. Rectangular boxes on left are products and pentagonal boxes on right are
operations performed on those products. In the Case Studies (see Sections 4, 5, 6
and 7), example SC criteria are derived using data from sites in a defined area that
have an ionic composition dominated by sulfate plus bicarbonate anions.
Cumulative distribution function (CDF), 95th centiie extirpation concentration
(XC95), 5th centiie hazardous concentration (HC05).
3.1.1. Establishing the Data Set
3.1.1.1. Information Sources
The data sets are developed at an ecoregional scale to account for natural differences in
background SC levels found in different ecoregions, and QA/QC of the data sets are described.
3-4

-------
Several data sets were used to illustrate the method in the case studies, details of which are
provided in Sections 4, 5, 6, and 7. In addition to the description of field data, information
available in the scientific literature is considered in the development of ecoregional SC criteria,
particularly for assessing causation and confounding factors (see for example Appendices A and
B in U.S. EPA, 201 la; Cormier et al., 2013b; Suter and Cormier, 2013) and to support
recommendations for duration and frequency of the criteria. Relevant literature related to the
characteristics of the receptor organisms, the ions of interest, and the potential confounding
agents are also considered, much of which is presented in Section 2 (Problem Formulation) of
this document and also in the precursor reports and manuscripts (U.S. EPA, 201 la; Cormier
et al., 2013b; Suter and Cormier, 2013).
3.1.1.2. Selection and Adequacy of Data Sets
Developing aquatic life criteria for SC using the XCD method requires a large data set
with certain characteristics. The adequacy of the data set can be judged by the following
attributes (U.S. EPA, 2011c):
•	Measurements of the agent(s) are paired in space and time with biological sampling;
•	High-quality (i.e., minimally affected) sites are included in the data set;
•	Background SC levels are similar throughout the region (see Section 3.7.1);
•	Characteristics of the agent (i.e., ionic composition) are similar across the region for the
paired data (i.e., other mixtures may occur but they are analyzed separately);
•	Some biological sampling occurs when salt-intolerant genera are likely to be collected
(e.g., March through June in Appalachia) and where they are likely to occur (e.g., leaf
packs, riffles);
•	The exposure gradient is broad enough to include no effects, weak effects, and strong
effects;
•	Data are available to evaluate potential confounding factors; and
•	An independent data set or statistical models are available to validate the criteria.
3-5

-------
Inclusion of many genera and a representative proportion of salt-intolerant genera help to
ensure that the XCD model is representative of the aquatic community. A sensitivity analysis of
data sets has determined that a reliable HCos can be determined from 90-120 genera and
500-800 sites, based on the stability of the HCos value as those variables increased
(U.S. EPA, 201 la; see Sections 4 and 5 of this report). Samples taken throughout the year
reduce biases from seasonal SC regimes and seasonal occurrence of some genera. For example,
samples taken only in dry seasons when SC tends to be higher would likely bias results toward
more salt-tolerant genera and to maximum SC exposures rather than an annual average.
3.1.1.2.1. Sample size
The number of observations of a genus can affect the reproducibility of the XC95 and the
HC05 (Cormier et al., 2013a). Similarly, the number of genera affects the reproducibility of the
HC05 (U.S. EPA, 201 la; Cormier et al., 2013a) and the number of genera depends on the overall
number of samples in the data set and individuals in the sample that are identified to genus.
For the example case studies, EPA estimated XC95 values using the XCD method with
genera that were observed in >25 samples in the ecoregion (Cormier et al., 2013a) because
estimations of the 95th centile with <25 observations are less robust. The recommended number
of genera in an XCD using this method is 90 (Cormier et al., 2013a). For a sampling protocol
that identifies 200 individuals in a sample, the adequate number of sampling stations in an
ecoregion using this method is about 500 (U.S. EPA, 201 la). However, if more individuals are
identified in each sample, e.g., 300 or a full count of individuals, then fewer sampling stations
may be needed to obtain XC95 values for 90 or more genera (see Appendix G in U.S. EPA,
2011).
The effect of selecting a minimum number of observations of a genus for calculating an
XC95 value can be visualized in several ways. For example, for a range of genus sample sizes
(e.g., N= 5-60), the HC05 is calculated and the number of genera in the XCD is enumerated.
The resulting HC05 values and number of genera versus minimum number of samples are plotted.
As taxa with fewer occurrences are excluded from the XCD, the number of genera decreases and
the HC05 increases (see Figures 4-11 and 5-11). To ensure representation of salt-intolerant taxa
and reasonable accuracy of the HC05, the minimum number of samples is chosen that maximizes
the number of taxa in the XCD while minimizing the variance in the XC95 and resulting XCD
3-6

-------
near the 5th centile. The number of observations of a genus is also chosen to provide sufficient
occurrences to estimate the genus XC95 values while minimizing bias due to eliminating
salt-intolerant genera that have few occurrences. A minimum sample size of 25 maximizes the
number of taxa that are included without having to extrapolate beyond the range of the data set
(Cormier et al. 2013a, on-line supplemental material). Therefore, a sample size of 25 was
utilized in the case studies, and in general, is a recommended minimum sample size for this
method.
The effect of sample size on the HC05 and its confidence bounds can be estimated using a
bootstrapping technique (see Section 4.5 and 5.5). Bootstrapping is a statistical technique of
repeated random sampling from an empirical data set that is often used in environmental studies
to estimate confidence limits of a parameter (Newman et al., 2001, 2000). This is akin to having
different samples to compare results and fidelity of the model. A similar method is used to
calculate confidence bounds on the HC05 values (described in Section 3.1.3.1). Using this
technique, a data set of a selected sample size with replacement is randomly selected from the
original set of samples. Next, the XC95 for each genus is calculated from the bootstrap data set
by the same method applied to the original data, and the HC05 is calculated. The uncertainty in
the HC05 value can be evaluated by repeating the random sampling and HC05 calculation
numerous times (e.g., 1,000 times) for each selection. The distribution of 1,000 HC05 values is
used to generate two-tailed 95% confidence bounds on these bootstrap-derived values. The
whole process is repeated for a selected sample size ranging from 100 to the full data set of all
samples. The mean of all bootstrapped HC05 values, the numbers of genera used for the HC05
calculation, and their 95% confidence bounds are plotted to show the effect of sample size. The
number of samples at which the HC05 values reach an approximate asymptote (500-800)
suggests the minimum sample size (500) for a data set (see Figures 4-12 and 5-12).
3.1.1.2.2. Treatment of multiple samples from a particular site
Multiple samples collected from the same site can provide valuable information
especially when they are from different seasons (when SC may be different and different genera
may be present), but they can be problematic if they introduce a bias (e.g., if extremely low or
high SC sites are more likely to be sampled repeatedly).
3-7

-------
In the case studies, most of the sites were sampled only once, but a portion of them were
sampled more than once (see Sections 4.1 and 5.1). To assess for a potential bias, a simple
inverse weighting scheme can be applied (e.g., if a site is sampled twice, each observation is
weighted 0.5). In the example case studies, this weighting scheme did not substantially change
the magnitude of the HCos values; therefore, all data were used with no weighting. In cases
where geographic distribution of sites and/or SC levels affects the HCos, then some form of
correction may be warranted, such as random selection of only one sample per site. In the case
studies, sites were fairly evenly distributed; therefore, no weighting was performed for replicates.
3.1.1.2.3.	Stressor identity
The stressor identity in this case is the proportion of constituent ions, characterized on the
basis of the field data set. The stressor of concern in the example case studies is an ionic mixture
dominated by sulfate (SO42 ) plus bicarbonate (HCO3 ) (see Section 2.2). As a result, for these
example criteria, sites with an ionic mixture dominated by chloride (i.e., those where the
concentration of HCO3 plus SO42 < CP, in mg/L) are removed from the data set. The ionic
mixture in the example case studies (see Section 4, 5 and 7) are dominated by the cations on a
mg/L basis by ([Mg2+] + [Ca2+]) > [Na+] and one by ([Mg2+] + [Ca2+]) < ([Na+] + [K+]) (see
Section 6). Alternatively, other data can be removed to focus on other mixtures or salts
(e.g., NaCl).
3.1.1.2.4.	Ambiguous taxa
The XCD method uses genus-level taxonomic identification. This method does not mix
data of lower or higher levels of taxonomic identification. However, species-level taxonomic
identification can be used when it is available and the number of species is sufficient for
constructing an XCD (see Appendix G for an example). Data records with ambiguous
taxonomic identification or family-level or higher identification (i.e., no genus-level
identifications) are excluded from the data sets.
3.1.1.2.5.	Exclusion of genera from extirpation concentration distribution (XCD)
This method is for freshwater systems, and therefore, estuarine and marine genera were
not included in the XCD. One way to ensure that only freshwater organisms are represented in
3-8

-------
the XCD is to only include genera that are present at a minimum of one freshwater reference site.
The selection of reference sites is beyond the scope of this document (Stoddard et al., 2006;
Herlihy et al., 2008; Whittier et al., 2007b; Hawkins, et al. 2010; U.S. EPA, 201 Id; Environment
Canada, 2012). For the example case studies, the reference sites used in analyses were identified
by the sampling organization, but only after EPA reviewed the reference site selection criteria
(see Sections 4 and 5). When reference sites were not identified by the sampling organization,
all genera were used (see Appendix D).
3.1.1.2.6. Confounding factors
Field observations are uncontrolled, largely unreplicated, and may not be randomized; as a
result, they are subject to confounding. Confounding is the appearance of apparently causal
relationships that are due to noncausal correlations. Noncausal correlations and the inherent
noisiness of environmental data can obscure true causal relationships. Reducing confounding as
much as possible is recommended by identifying potential confounding variables; determining
their contributions, if any, to the relationships of interest; and eliminating their influence when
possible and as appropriate based on credible and objective scientific reasoning.
A method to assess the potential effect of confounders is described in the EPA
Benchmark Report (U.S. EPA, 201 la) and in Suter and Cormier (2013). The analysis of
potential effects on the model by potential confounders used a weighted scoring system to
evaluate ten types of evidence that determined whether the observed relationship between
benthic macroinvertebrate community composition and SC was affected by other factors. The
weighted scoring system was based on work by Hill (1965) and Cormier et al. (2010) and is
described in detail in Appendix B of EPA (U.S. EPA, 201 la) and in Suter and Cormier (2013).
As described in the EPA Benchmark Report, the potential for other stressors to affect the XCD
model was evaluated using a weight-of-evidence assessment that considered habitat quality,
organic enrichment, nutrients, deposited sediment, pH (low and high), selenium, temperature,
lack of headwaters, catchment area, settling ponds, dissolved oxygen, and metals (see
Appendix B in U.S. EPA, 201 la). Overall, the analyses showed that the effects attributed to
increased ionic concentration were not due to other stressors.
In these analyses (U.S. EPA, 201 la), only one of the assessed factors (pH <6) was
identified as a likely confounder, so samples with pH <6 were removed from the data set to
3-9

-------
minimize the influence of acidity and associated dissolved metals. Due to the toxicity at low pH,
unless shown to the contrary, sites with pH <6 are excluded from data sets prior to analysis.
Also, when using this method, the EPA recommends that at least one sequence of
analyses be used to evaluate the effect of confounding on the XCD model. For example,
potential confounding can be evaluated using multiple regression analyses followed by
modification of the data set to control for the strongest potential confounding factors and then
calculating the HCos. If the HCos is not appreciably altered by this data set manipulation, the full
data set can be used. If the HCos is appreciably altered, the criterion data set may be modified to
minimize that factor's effect on model prediction. The model is accepted if the confidence
bounds of the original and new HCos overlap. Examples for confounding analyses and reducing
effects of confounding can be found in Case Studies I and II in Appendices A.2 and B.2,
respectively.
There are two common means for reducing the influence of confounders. First, sites with
a confounder can be removed from the data set, thus reducing its influence on the XC95 estimates
and XCD model. For example, the EPA removed samples with low pH in the case study
examples (see Appendices A.2.3 and B.2.3). Secondly, the effect of a confounder can be
minimized by normalizing the influence of a confounder with appropriate weighting. The EPA
used this approach to assess the influence of temperature and season in the case study examples
and these methods could be used to adjust for confounders if necessary (see Section 3.1.4, and
Appendices A.2.3 and B.2.3). Removing samples from the data set can reduce the number of
species or the range of exposures of the stressor of interest, thus affecting the reliability of the
estimates. Therefore, it is important to evaluate whether the manipulation of the data set
improves the accuracy of the HCos. Each case is different, and professional judgment is
recommended.
3.1.1.3. Quality Assurance/Quality Control (QA/QC)
Information is reported about the specific methods used to choose sampling locations, to
sample water SC and macroinvertebrates, and to assure data quality. Some considerations
include whether standardized quantitative or semiquantitative techniques are used for
macroinvertebrate sampling, the mesh size used in the field and lab for sampling and sorting,
whether samples are subsampled, and if so, what percentage was subsampled. The data set
3-10

-------
description and metadata include sampling dates; total number of chemical, physical, and
biological samples from distinct locations (and total samples); sampled years; stream types
represented by samples; and ecoregions represented by sites. Annual sampling when
salt-intolerant genera are likely to be present is usually sufficient and avoids damage to the
habitat and stream biota due to repeated sampling during the year. If sampling occurs more than
once per year from the same location, the use or restriction of repeat samples is described.
Similarly, a full description or literature references are provided for the chemical and physical
parameters that are used in any analysis (see an example in Section 4.1). Additional information
regarding QA/QC and other critical technical elements of a robust biological assessment program
(e.g., taxonomic resolution, sample collection, sample representativeness, sample processing,
data management, and professional review) can be found in EPA's technical assistance
document, Biological Assessment Program Review: Assessing Level of Technical Rigor to
Support Water Quality Management (U.S. EPA, 2013a).
3.1.2. Calculating Genus Extirpation Concentrations (XC95)
For each genus meeting the data-selection conditions, a CDF is constructed that is
weighted to correct for any potential bias from the unequal distribution of sampling of sites
across the range of logarithm 10 transformed SC values. This weighted CDF represents the
proportion of observations of a genus with respect to increasing exposure levels. The extirpation
effect threshold for a genus is 95% of the total occurrences of the genus. The two exposure
levels bracketing the 95th centile are linearly interpolated to give an XC95 for a genus.
In the case examples, all calculations are performed using logarithm base 10 (log 10)
transformed SC values. Variables are routinely log transformed when applying the field-based
methods. Because environmental data are usually skewed, log transformation normalizes the
data so that normality assumptions are not violated. Log transformation also tends to increase
equality of variance, increase the linearity of relationships, and makes plotted relationships
clearer.
First, equally-sized bins are defined to compute weights for each sample. Bin size
depends on the data set and is based on balancing the requirements of sufficient observations in a
single bin to define the proportion and a sufficient number of bins to define the form of the
response. The effect of bin size can be analyzed by developing a series of HC05 values using
3-11

-------
different bin and sample sizes. In general, 40 to 60 bins usually gives acceptable results. For
example, for Case Studies I and II, the width of each bin is l/60th of the range of the loglO
transformed SC values. Thus, each bin was assigned a width equal to 0.017 (1/60 bins)
multiplied by the range of the loglO transformed SC values within the data set (for examples see
Figures 4-6 and 5-6).
Next, the bins are weighted to ensure that sites in bins with many observations are not
overly influential. The assigned weight for each sample within a given bin is Wi = 1 ///,, where n,
is the number of samples in the z'th bin. The value of the weighted cumulative distribution
function, F(x), is computed using the following equation for each unique observed value of the
agent x associated with observations of a particular genus:
ZH'
-------
genera). For an example of an XCD cumulative distribution plot of XC95 and HC05 derivation,
see Figure 4-7.
3.1.3.1. Validating the Effect Estimate Hazardous Concentration (HCos) by Bootstrapping
The EPA's Science Advisory Board (U.S. EPA, 201 lc) recommended using
bootstrapping as one way to validate the XCD model and estimate uncertainty around the HCos
values. This method generates distributions and confidence bounds for each genus in the first
step and propagates the statistical uncertainty of the first step through the later steps in which the
XCD is created and the HCos is estimated (see Figure 3-5). A data set of the same sample size as
the original data set is randomly selected with replacement from the original set of samples
(Efron and Tibshirani, 1993). The XC95 for each genus that occurs at least once in the original
data set's reference sites and in more than 24 sampled sites is calculated from the bootstrap data
set by the same method applied to the original data. The XC95 for each genus is stored and later
used to estimate the confidence bounds of each genus' XC95. Then, the HCos is calculated and
stored. The uncertainties in the XC95 and HCos values are estimated by repeating the sampling
and calculations 1,000 times. The distribution of 1,000 HCos values is used to generate
two-tailed 95% confidence bounds on these bootstrap-derived values (see Figure 4-13 for an
example). The particular genera and the number of genera in any bootstrapped XCD differ in
each bootstrapped sample of sites; therefore, the number of bootstrapped XC95 values of genera
may be more or less than the number of genera in the original data set. The distribution of
1,000 XC95 values for each genus is also used to generate two-tailed 95% confidence bounds on
these bootstrap-derived XC95 values. See Appendices A.3 and B.3 for example 95% confidence
bounds for each genus.
3-16

-------
Repeat 1,000 times
Store
results
Original 1,000
observations
Compute
XC95 and
HC05
values
Calculate 95%
confidence
bounds
Select
reference taxa
that occurred at
25 or more
samples
Randomly pick
1,000 samples
from observations
with replacement
Figure 3-5. Diagram from EPA Benchmark Report depicting the process for
estimating uncertainty. Bootstrapping is a resampling of the original data set to
create 1,000 new data sets and from each of the 1,000 data sets, the extirpation
concentration (XC95) values are calculated. After each run, an extirpation
concentration distribution (XCD) is made and a hazardous concentration (HC05)
estimated. This process is repeated until there are 1,000 HC05 values and then the
confidence limits of the HC05 are estimated. The number of samples varies
depending on the data set. The same resampling process is used to evaluate the
effect of different sample sizes, exclusion of genera using different database
selection criteria, or other parameter choices (see Section 3.1.1.2).
Confidence bounds represent the potential range of HC05 values using the XCD approach
based on the data set. Conceptually, these confidence bounds may be thought to represent the
potential range of HC05 values that one might obtain by returning to the field and resampling the
same set of streams. The contributors to this uncertainty include measurement variance in
determining SC and sampling variance in the locations for monitoring, collecting, and
enumerating organisms. These also include variance due to differences in stream reaches,
weather, and other random factors.
The confidence bounds do not address potential systematic sources of variance such as
differences in sampling protocols. The contributions of those sources of uncertainty, in addition
to the sampling uncertainty, can best be evaluated by comparing the results of independent
studies. One estimate of that uncertainty may be provided by comparing the all-year HC05 values
derived from the region for which criteria is being derived to another comparable region. Even if
data are obtained in different areas by different agencies using different laboratory processing
protocols, the HC05 values may be similar.
In the EPA Benchmark Report, the HC05 value was validated by an independent data set
which had a similar background SC, and the values differed by less than 5% (see Appendix G in
U.S. EPA, 201 la). Large data sets for ecoregions from more than one sampling organization are
3-17

-------
rare and so EPA has provided this bootstrapping method (Newman et al., 2000) as an alternative
method to validate the XCD model as suggested by the SAB in their review of the EPA
Benchmark Report (U.S. EPA, 201 lc).
3.1.4. Assessing Seasonality, Life History, and Sampling Methods
The seasonality of life-history events such as emergence of aquatic insects can affect the
probability of detecting a species because eggs and early instars are not collected by most
sampling methods. As an illustration, in the example cases in Sections 4 and 5, annual insects
(univoltine) that emerge in the spring, although present, are less likely to be detected in the
summer, when coincidently, SC levels increase in some streams (e.g., due to decreased flow). In
other locations, this pattern may be different. For example, high mountain systems may be
affected by melting snow pack. Seasons may shift based on latitude. Also, sampling restricted
to a season can bias the estimate of natural background or effect estimates (i.e., XC95 values).
Both high-concentration and low-concentration periods should be represented when the
salt-intolerant genera are collected in order to ensure that the tolerated range is evaluated. These
periods may vary by time of year among regions, and among years (based on climate variability)
for any given region. Professional experience with the SC regimes and the life cycles of
vulnerable species is required when assessing whether a data set is suitable for using the XCD
method.
Because the hydrologic and SC regime and the natural history of salt-intolerant taxa vary
by region, the potential effect of sampling date on the form of the XCD model may be assessed
using several methods. As an illustration, in the example cases in Sections 4 and 5, the HC05
using the spring (March-June) only data set was compared with the HC05 based on the full-year
data set for the ecoregion. If the spring HC05 is within the confidence bounds of the full-year
data set, then the full annual data set can be used (see example in Appendix A, Figure A-8); if
this is not the case, further correction for a confounding factor may be necessary as described
below for the example using sampling date.
A scatter plot and regression model can be developed to evaluate the mean relationship
between measurements of SC at the time of the biological sample and annual mean SC (for an
example, see Appendix A, Figure A-9). The annual geometric mean SC values are calculated
from at least six water samples collected before biological samples were taken. At least one
3-18

-------
spring (when salt-intolerant taxa can be collected) and one summer macroinvertebrate sample is
recommended in order to increase the likelihood that salt-intolerant taxa will be included in the
data set. On the x-axis is the SC when biological samples are collected and on the_y-axis is the
annual geometric mean value during that rotating year for a site. A Model II regression is fitted
to the data because the two SC measurements are uncertain. [A Model I regression using least
squares is not used because it underestimates the slope of the linear relationship between the
variables when they both contain error (Legendre, 1998)]. The mean relationship between
measurements of SC at the time of the biological sample and annual mean SC is supported if the
relationship approaches 1:1, and prediction for the annual mean from the regression model is
within the confidence bounds of the HCos.
A third approach to account for seasonal variability involves adjusting SC results
collected at the time of biological sampling to estimate annual mean SC values. In situations
where the SC tends to be lower in the spring than in the summer, the effect of seasonality on the
HCos is evaluated by converting the instantaneous biological sample SC into an annual mean
value based on monthly weighting factors. Then, the XC95 and HCos values are estimated. The
average weighting factors for each month can be calculated from the previous sampling year.
One way to do this is to select a subset of stations where multiple SC measurements are taken
within a rotation year (e.g., from July to the following June). For each site, the annual mean SC
is calculated using the monthly measurements. Then, the weights for each month are calculated
as a ratio of the annual mean SC to the observed SC at each site on the day of biological
sampling. Next, the average weight within each month is calculated. Finally, for the data set
used to develop the XCD and the HCos, the SC on the day the biological sample was collected is
multiplied by the weighting factor for that month to yield the estimated annual SC for each site.
The resulting products are considered the annual mean SC at each station in that rotation year,
adjusted by month. The weighting factors vary slightly for different months in different data sets
(see Appendix A, Table A-2 for an example). This approach may be adapted to different
seasonal SC patterns, as appropriate.
If the confidence bounds of the weighted HCos overlap the confidence bounds of the
unweighted HCos, the unweighted model is accepted. For example, the HCos values in Case
Studies I and II vary by less than 3%, suggesting that the impact of sampling date (seasonality) is
minor for these data sets. As a result, seasonally unweighted XCDs are used for the assessment
3-19

-------
in these case studies. In general, the use of unweighted XCDs is easier and requires fewer data
points. However, all data sets may not yield similar results, and thus this method provides a way
to evaluate the influence of season and also provides a method for normalizing for sampling date
when necessary.
3.2. DERIVING A CRITERION MAXIMUM EXPOSURE CONCENTRATION (CMEC)
A CMEC is defined as the SC level that protects aquatic life from acutely toxic
exposures. The CMEC analysis described here estimates the 90th centile of observations at sites
with water chemistry regimes for sites meeting the CCC. It is not directly estimated from paired
biological and water chemistry during acute exposures. However, if sufficient data are available
(e.g., daily measurements of SC paired with macroinvertebrate sampling), a protective criterion
maximum concentration could be estimated from the maximum concentration in a year prior to
the observation of salt-intolerant genera at a site. An example of this type of analysis is provided
in Appendices A.3 and B.3, but such data sets are rare. Even for the case studies using the large
data sets in Ecoregions 69 and 70, there are only modest amounts of data to estimate such a
criterion maximum concentration in this manner.
Using only water chemistry measurements and a previously determined CCC, a CMEC is
estimated such that where the CCC is attained, 90% of observations are likely to be less than the
CMEC. The steps involved in selecting, characterizing, and analyzing SC (chemistry) sampling
data to derive a CMEC for flowing waters in the study area are described below (see Figure 3-6)
and example derivations in Sections 4.2.2 and 5.2.2.
3-20

-------
Field Data Set
Develop
Data Set
Data set of sites with frequent
resampling of the stressor
Include sites with stressor, (e.g., SC)
sampled >6 times
Check
unequal
variance
Estimate
type 1 error,
a = 10%
Calculate
CMEC
Calculate within site variability
Verify that the variation is stable near
the CCC, then select final data set
Calculate standard error,
then calculate CMEC
Set of sites with annual
mean inclusive of CCC value
90th centile of observations for
sites meeting CCC
Scatter plot and LOESS
regression line with CI

Criterion Maximum Exposure
Concentration
Round to 2 significant figures
Figure 3-6. Main steps in the derivation of a criterion maximum exposure
concentration (CMEC) based on field water chemistry data. Rectangular
boxes on left are products and pentagonal boxes on right are operations performed
on those products. A locally weighted scatterplot smoothing (LOWESS)
estimates a line for a scatter plot by iteratively calculating many nonparametric
regression models using local approximations (linear polynomial) from
neighboring points.
The CCC is expressed as an annual geometric mean of the SC values measured from a
sampling station in a particular region. Because SC values vary spatially and temporally, it is
expected that the maximum SC values (or the CMEC) at any given station may be estimated by
incorporating both among-station (spatial), and within-station (temporal) variability. Using the
mean SC at the CCC and the variance of a SC distribution, a centile near the maximum, such as
the 90th centile, for that distribution can be estimated. Thus, where the CCC is met within the
region, only 10% of the observations (grab samples) would be predicted to exceed the CMEC.
3-21

-------
To do this, a subset of frequently sampled sites is developed. For Case Studies I and II, a
representative sample set was identified in which n stations (j in 1,.., n stations) were sampled at
least six times (sample size fcj, ith ml,..., kt observations) on a rotating yearly basis (in the case
example from July to June). The preferred data set would have multiple SC measurements
evenly distributed throughout the year. A minimum of one sample during the low SC season
(e.g., March-June in Appalachia), and one sample in the high SC season (e.g., July-October in
Appalachia) may be sufficient to capture temporal variability. As with the derivation of the
CCC, a range of exposures that leads to adverse effects on the most salt-intolerant taxa needs to
be represented in the data set and there needs to be assurance that there is no bias in the sampling
within that range. The grand mean and standard deviation of this data set are calculated. The
CMEC can be calculated at the 90th centile of the distribution from log values of SC in the region
from this equation:
CMEC = 10^+z«*^)	(3-2)
Where X is the proposed annual geometric mean value limit for all stations (x,) (i.e., the
CCC), za is the one-tail critical value for the 90th centile of a normal distribution (a, 10%), ar is
the total residual standard deviation, i.e., the square root of the standard deviation. The CMEC is
calculated based on eq 3-2 with X equal to loglO of the CCC for the ecoregion. For example, if
the grand mean of all sites is 310 [j,S/cm, and the standard deviation is 0.243 [j,S/cm, and the za at
90th centile is 1.28, then the estimate of 634.1 [j,S/cm is rounded to two significant figures
resulting in a CMEC of 630 [j,S/cm (see eq 3-3).
10logl0(310) + 1.28*0.243 = 63Q ^s/cm	(3.3)
3.3. ESTIMATION OF CRITERIA DURATION
The water quality standards handbook (U.S. EPA, 1983) describes duration as follows:
The quality of ambient water typically varies in response to variations of effluent
quality, stream flow, and other factors. Organisms in the receiving water are not
experiencing constant, steady exposure but rather are experiencing fluctuating
3-22

-------
exposures, including periods of high concentrations, which may have adverse
effects. Thus, the EPA's criteria indicate a time period over which exposure is to
be averaged, as well as an upper limit on the average concentration, thereby
limiting the duration of exposure to elevated concentrations.
Because this field-based approach relies directly on paired in situ measurements of SC
and benthic invertebrate assemblage composition, the potential adverse effects of ionic stress on
all life stages are considered in the context of other complex relationships (e.g., food web
dynamics) and aquatic ecosystem processes. The measures of effect (i.e., XC95 and HC05) are
considered chronic-duration endpoints because the field data reflect exposures over whole life
cycles and multiple generations of the resident biota (see Table 2-3).
The EPA typically recommends an averaging period of 4 days for a CCC, which may be
appropriate for some field-derived criteria (U.S. EPA, 1985). Important considerations for
estimating duration are the temporal resolution of the biological data set and the seasonal
window for observing salt-intolerant genera (typically early in the year). Based on available
field data, salt-intolerant macroinvertebrate genera may be exposed to a range of SC levels
greater than the CCC throughout the year and often for more than 4 days (see example in
Figure 3-7). For example, biological samples collected once annually (as in Case Studies I and II
in Sections 4 and 5) represent the average stream chemistry and macroinvertebrate assemblage
information over the course of 1 year. In cases where samples were collected on an annual basis
the EPA recommends a duration of 1 year for CCCs for SC derived using the XCD method
unless there are sufficient data to support an alternative duration.
3-23

-------
OT-00001-8.8
2000 2002 2004 2006 2008 2010
Figure 3-7. Typical specific conductivity (SC) pattern of a stream with
annual mean SC well below 310 jiS/cm. Aquatic life is typically exposed to a
range of SC levels throughout the year.
The duration for the CMEC in Case Studies I and II is based on a literature review of the
rate of onset of critical biological responses and the sampling duration used in the field data set
used to establish the CMEC. Although reproductive effects (see Section 2.5) may occur rapidly
following exposure, they occur only during distinct temporal windows that vary with species
(life history). Increased drift (benthic invertebrates floating downstream), in contrast, can occur
any time a spike in exposure occurs. In numerous studies, increased drift may be induced within
minutes of stressful exposures in streams and in artificial test channels (Svendsen et al., 2004;
Wood and Dykes, 2002). Most ecological studies describe drift as a part of the natural history of
dispersal and colonization, but disturbance has also been identified as a cause of drift (Svendsen
et al., 2004; Wood and Dykes, 2002; Crossland et al., 1991; Doeg and Millage, 1991; Wallace,
1990; Brittain and Eikeland, 1988; Sheehan and Winner, 1984; Geckler et al., 1976; Waters,
1995, 1972, 1966). In a study that induced drift from the addition of sodium chloride (NaCl), the
onset of drift occurred within 15 minutes, and on average, the greatest occurrence of drifting
genera took place within 4 hours (Wood and Dykes, 2002). In that study, prior to the addition of
salt, SC was 110 [j,S/cm (River Holmes). During three trials, drift occurred at maximum total
dissolved solids (-110 mg/1 or SC -157 (j,S/cm) and not at the lower concentration (-85 mg/L or
SC -121 (j,S/cm). In stream mesocosm studies with HCO3 and SO42 salts, Clements and
3-24

-------
Kotalik (2016) reported an approximate doubling of drift within 24 hrs (196 |iS/cm), Avoidance
and drift behaviors, in general, typically occur quickly following most noxious stimuli.
Unlike laboratory studies, maximally tolerated SC exposures were not measured. Rather,
the CMEC is the 90th centile of observations measured at sites meeting the CCC. Therefore, the
duration parameter is based on the rate of onset of drift described in published field experiments;
the recommended duration for the CMEC is 1 day.
3.3.1. Summary of Recommended Duration for Criterion Continuous Concentration
(CCC) and Criterion Maximum Exposure Concentration (CMEC)
The temporal resolution of the biological data set and the seasonal window for observing
salt-intolerant genera are key considerations when estimating the duration for the CCC and
CMEC for SC. In cases where sampling occurs once annually, as in Case Studies I and II
provided in Sections 4 and 5, the recommended durations for the CCC and CMEC are 1 year and
1 day, respectively.
3.4. ESTIMATION OF CRITERIA FREQUENCY
The water quality standards handbook (U.S. EPA, 1983) describes frequency as follows:
To predict or ascertain the attainment of criteria, it is necessary to specify the
allowable frequency for exceeding the criteria. This is because it is statistically
impossible to project that criteria will never be exceeded. As ecological
communities are naturally subjected to a series of stresses, the allowable
frequency of pollutant stress may be set at a value that does not significantly
increase the frequency or severity of all stresses combined.
The frequency with which criteria may be exceeded depends on the rate of recovery of
the biotic community. In general, if the interval between exceedances is less than the time to
recovery, impairment is perpetuated. Time to recovery may be estimated mechanistically from
the life histories of the organisms involved or empirically from field studies of stream
community recovery.
In this case, to estimate the interval between extirpation and reestablishment of a
reproducing population, the EPA adapted a list of potential factors that may affect recovery time
of stream organisms (Wallace, 1990). Although these considerations were originally outlined
3-25

-------
with respect to pesticide exposure, they are a reasonable list of factors to consider for any
stressor. The revised considerations are listed below with corresponding numbers from Wallace
(1990):
la. Magnitude of the initial stressor load
lb. Adverse effect of the stressor
lc. Areal extent of continued inputs of the stressor
2.	Spatial scale of the disturbance
3.	Persistence of the stressor at the site
4.	Vagility of the populations (ability to move and disperse) influenced by exposure
5.	Timing of contamination in relation to life history stage
6.	Position within the drainage network
For the purposes of estimating frequency, the concentration of the initial stressor load
(see Item la) is greater than the magnitude of an annual average CCC or a daily CMEC. The
initial adverse effects (see Item lb) are extirpation resulting from drift and failure to recruit.
Because insect colonization (the predominant invertebrate group in streams) is sometimes
possible via aerial dispersal, frequency was estimated for conditions where the effect of areal
extent (see Item lc) or spatial scale of the disturbance (see Item 2) is minimal. However, on a
case-by-case basis, if the disturbance by ionic concentration is spatially extensive, the frequency
recommendation for these criteria might not be protective (Smith et al., 2009; Lindberg et al.,
2011; Bernhardt et al., 2012). Wallace (1990) defines persistence (see Item 3) as continuation of
exposure in the environment after cessation of new inputs. Because salts are highly soluble in
water, they are flushed downstream (in flowing waters) when loading stops, and therefore, they
are not persistent chemicals in the sense defined by Wallace (1990) although lag times can be
long when contaminated groundwater flows to surface waters. Although intermittent releases of
water with high specific SC result in intermittent exposures, the aquatic impacts appear to be
long term owing to persistent exposures (U.S. EPA, 201 lb; Pond et al., 2014; Evans et al., 2014;
Williams, 1996; Feminella, 1996; Delucchi, 1988; del Rosario andResh, 2000). Recolonization
rates from upstream sources or connected tributaries that provide a source of drifting juveniles
3-26

-------
was not used to estimate frequency so as to provide a conservative estimate and because in some
situations, such as headwater streams, there may not be an upstream source of recolonizing
juveniles (see Item 6). Using these conditions, the life history of salt-intolerant benthic
invertebrates (see Item 5) is considered with respect to recolonization potential (vagility) (see
Item 4) from aerial dispersal of adults and oviposition.
3.4.1. Recovery Rates in Literature Reviews
The frequency parameter for this method is estimated from ecosystem recovery rates
following disturbance as reported in the literature. In this case, frequency is an estimate of the
period of time between macroinvertebrate extirpation and recovery (reestablishment) of the
population. The estimate of time to recovery is based on life cycles and natural history. The
assessment is supported by a literature review of the recovery of aquatic macroinvertebrates
following chemical and nonchemical-induced effects in 31 nonflowing systems (lentic) and
111 flowing (lotic) systems reviewed by Niemi et al. (1990) and more than 12 streams reviewed
by Wallace (1990). Niemi et al. (1990) indicated that recovery time was less than 3 years except
when (1) the disturbance resulted in physical alteration of the existing habitat, (2) residual
pollutants remained in the system, or (3) the system was isolated and recolonization was
suppressed. The frequency estimated for SC criteria applies when the three factors listed above
are not operative; that is, physical alteration has not taken place, pollutants are flushed from the
system, and colonization is possible.
Ionic regimes may be long lasting. In a study of 15 valley fills, Pond et al. (2014) found
that SC remained elevated 11-33 years after reclamation. Despite good instream habitat, nearly
90% of these streams exhibited biological impairment. Valley fill sites with higher index scores
were near unaffected tributaries, an indication that drifting colonists accounted for the presence
of sensitive taxa. Based on 137 valley fills, Evans et al. (2014) estimated that it would take
approximately 20 years to potentially attain SC levels <500 [j,S/cm after initiation of valley-fill
construction. These two studies underscore the fact that although recovery can occur within
3 years when the exposure no longer exists, some streams may take decades to return to levels
that salt-intolerant genera can tolerate and maintain viable populations.
Wallace (1990) also indicated that the definition of recovery was inconsistently applied in
the scientific literature and in most cases true recovery was not attained within the study interval;
3-27

-------
that is, some species had recolonized but the original assemblage was altered. Many other
studies of invertebrate recovery times from single events, such as toxic chemical spills, floods,
pesticides, drought, and organic pollution, indicate recovery times of a few months to several
years (e.g., Kattwinkel et al., 2012; Molles, 1985; Minshall et al., 1983; Heckman, 1983; Fisher
et al., 1982; Hynes, 1960; Mebane et al., 2015). However, in some situations, biological
communities may have faster recovery times (Wood and Dykes, 2002) if the exposure duration is
short and if there are upstream sources for recolonization. In a more recent review of 200 studies
with pesticide exposures, Kattwinkel et al. (2012) reported that migration from upstream
uncontaminated areas is a main driver for recovery and that recolonization varied with
generation time and source of migrants; however, upstream sources of colonizers may not be
present (e.g., in headwater streams). Faster recovery times were related to drift from external
sources (Caquet et al., 2007; Liess and Schulz, 1999) or untreated refugia (Brock et al., 2009).
Salt-intolerant univoltine species (life cycles of 1 year) do not recover within 1 year after
exposure (Liess and Beketov, 2011; Liess and Schulz, 1999). Overall, most studies reported that
recovery took two or more generations and as many as five generations even with upstream
sources that can recolonize by drift. Furthermore, there is a trend of longer population recovery
time with increasing generation time; species with long generation cycles often take longer for
population recovery.
Most of the macroinvertebrate genera sensitive to ionic stress have a univoltine
generation time (1-year life cycle), and therefore, their recovery time is likely to be longer
compared to multivoltine genera (having less than 1-year life cycles). Where recolonization by
juveniles drifting from upstream refugia is not possible, aerial dispersal from nearby streams
would be necessary to reestablish populations of aquatic insects; in this case, recovery may take
longer than 3 years. The frequency recommended for this method assumes there are either
sources from upstream or airborn dispersal for recolonization.
In summary, if the concentration of major ions in a stream can be returned to levels that
are capable of supporting aquatic life, and if the physical habitat is suitable, and if there are
opportunities for recolonization from an upstream source and/or through aerial dispersal from
nearby streams, then the allowable recommended frequency of exceedance for criteria derived
using this method is once every 3 years. If any of these conditions are not met, then the
3-28

-------
frequency parameter is expected to be more than 3 years, because the stream lacks the ability to
recover within that span of time.
3.4.2. Life History Considerations
Often, more than 90% of benthic invertebrates in streams are insects. Several of the most
salt-intolerant benthic insects (e.g., mayflies, stoneflies, caddisflies, and true flies) have a
1-2 year life cycle with emergence, mating, and early development occurring in the spring
months (Merritt and Cummins, 1996; Brittain, 1982; Clifford, 1982). Hypothetically, if a
univoltine genus is extirpated in the first year and in the following spring migrating insects laid
eggs, offspring from the colonizers would be large enough to be observed in the collections the
following year (i.e., 2 years after the initial extirpation event). Assuming that a recovered
population required two reproductive seasons (Liess and Beketov, 2011; Liess and Schulz,
1999), the earliest measurable recovery would be the year after that, or 3 years after the initial
extirpation event. The genetic diversity of the population founded by a few colonists may be low
(i.e., the founder effect) and as a result the population may be less resilient.
Gastropods, amphipods, isopods, and crayfish tend to be more tolerant to ionic stress (see
Appendix E); these taxa would likely remain provided that SC levels did not exceed their
predicted extirpation concentrations. Extirpation of most noninsect benthic invertebrates is not
expected to occur if the yearly average is <960 [j,S/cm, the XC95 of the most salt-intolerant
crustacean based on values calculated using a combined data set from Case Studies I and II
(U.S. EPA, 201 la, see Appendix D). The natural history of fish suggests that they may be able
to recolonize quickly due to their greater mobility; however, immigration may be limited for
some species because they are endemic to specific drainages or there may be barriers to
emigration (Hitt and Chambers, 2014; see Appendix G). Unionid mussels were not evaluated by
the EPA, but some field and laboratory studies suggest that Unionidae are also salt-intolerant
(Price et al., 2014; Gillis, 2011; Wang et al., 2013, Kunz et al., 2013). If immigration of fish is
restricted or if less mobile species such as mollusks and crayfish are extirpated, their
recolonization could take much longer than 3 years, or may require reestablishing a colonizing
population by stocking (Wallace, 1990).
3-29

-------
3.4.3. Summary for Field-Based Frequency for Criterion Continuous Concentration
(CCC) and Criterion Maximum Exposure Concentration (CMEC)
If SC criteria derived using this method are exceeded, it is expected that at least 5% of
macroinvertebrate genera will be extirpated, and many more genera will have been exposed to
levels that reduce their occurrence (U.S. EPA, 201 la). However, recovery is expected to occur
in 3 years if the following conditions are met: (1) the SC regime returns to a yearly average
below the CCC, (2) there are nearby streams with low SC supporting a diverse community, and
(3) there is an upstream source of colonizers or the flight or recolonizing distance is within the
dispersal range of genetically diverse, reproducing adult colonizers. This frequency
recommendation is based on consideration of the life history of insects that are able to recolonize
a site by drifting from upstream sites or aerially dispersing from a nearby stream, and published
studies of recovery of stream communities. If any of these conditions are not met, the time
necessary for community recovery (and thus, the allowable frequency of exceedance) would
likely be longer than 3 years.
3.5. ASSESSING CAUSATION
Field studies can generate statistical relationships between environmental attributes and
biological responses, but those relationships are not necessarily causal. Epidemiologists evaluate
whether an apparent relationship is causal by weighing evidence of causation in terms of lists of
considerations (Norton et al., 2015). General causation between SC and macroinvertebrate
occurrences was previously assessed (U.S. EPA, 201 la; Cormier et al., 2013b) and therefore
does not need to be repeated. Many other studies have corroborated that assessment for the
particular ionic mixture, Ca2+, Mg2+, HCO3 and SO42 , and for different salt mixtures, e.g., Na+,
K+, HCO3 and SO42 , and Na+ and CP (see Section 2.2). A new general causal assessment is
recommended when it is uncertain whether an agent, for example a newly synthesized chemical
or novel mixture, can or has ever harmed aquatic life. If a new causal assessment is warranted,
EPA recommends using epidemiological methods to demonstrate that the agent or mixture can
and does cause extirpation at concentrations using the method described in the EPA Benchmark
Report (U.S. EPA, 201 la; Cormier and Suter, 2013b). The causal assessment methodology does
not compare the relative importance of ionic-induced impairment with other known stressors
such as metal toxicity, stream bed erosion and siltation, or eutrophication (U.S. EPA, 201 lb;
Gerritsen et al., 2010). Effects from these stressors are likely to occur and do occur in any given
3-30

-------
ecoregion and at times concurrently with increased ionic inputs. Rather, the causal assessment is
designed to determine whether the addition of ions to streams can and does cause extirpation of
aquatic life.
Although causal assessments of most ionic mixtures do not need to be repeated, it is good
practice to evaluate the predictive performance of the XCD model, that is, how well the model
characterizes the modeled relationship, e.g., SC and extirpation. See Section 3.1.1.2.6 for
analytical approaches for assessing potential confounders.
In summary, the causal relationship between elevated ionic concentration and extirpation
of macroinvertebrates can be assessed using an approach modified from Hill's (1965)
considerations (for complete details see Appendix A in U.S. EPA, 201 la; Cormier et al., 2013b).
Hill's approach for establishing a probable causal relationship has been adapted for ecological
applications (Cormier et al., 2010; Fox, 1991; Suter et al., 2001; U.S. EPA, 2000b). Based on
that assessment, the body of evidence indicates that the loss of macroinvertebrate genera occurs
where SC is high even when potentially confounding causes are low, but is rare when SC is low.
Furthermore, there are sources of ions that increase stream SC in the region, and aquatic
organisms are directly exposed to these ions. Physiological laboratory studies indicate that ionic
gradients in high SC streams would not favor the exchange of ions across gill epithelia and that
physiological functions of organisms are affected by elevated ionic concentration. Some genera,
composite metrics, and assemblages are affected at sites with higher SC, while others are not.
Laboratory studies using moderately salt-intolerant species and ionic compositions relevant to
the study area support ionic stress as a cause of extirpation; and increased exposure to ionic
stress affects macroinvertebrate abundance and diversity based on field observations. More
recently, mesocosm studies have corroborated adverse effects at similar exposures (Clements and
Kotalik, 2016).
The causal assessment confirmed that the mixture of ions in streams with elevated SC
and neutral or somewhat alkaline waters can and is causing the extirpation of salt-intolerant
genera of macroinvertebrates as well as in low pH systems (U.S. EPA, 201 la; Cormier et al.,
2013b). The relative SC level of waters with a similar ionic composition, rather than any
individual constituent of the mixture, is implicated as the cause of impairment (see
Section 2.5.1). The causal relationship describes how Ephemeroptera and similarly
salt-intolerant invertebrates, in general, respond to ionic stress and does not require that the
3-31

-------
species or genera be the same in all applications or at all locations. Although the specific
constituents of the ionic mixture were not individually assessed, the cause of impairment is
attributable to one or more of the primary constituents of the mixture. Therefore, based on the
causal assessment (U.S. EPA, 201 la), it is expected that SC levels sufficient to cause
extirpations would occur with a similar salt mixture containing HCO3 , SO42 , Ca2+, and Mg2+ in
other regions. Furthermore, based on other studies (see Section 2.2), different salt mixtures also
cause extirpation, e.g., Na+, K+, HCO3 and SO42 , and Na+ and CP.
3.6. ASSESSING WATERBODY APPLICABILITY
3.6.1.	Stream pH
Due to the nature of ions and pH, it is important to consider the potential impact of pH on
the XCD. Acidity (e.g., associated with acid mine drainage, atmospheric deposition and other
sources) and potentially associated dissolved metals could affect the field-based XC95 values and
the XCD model. As a result, unless shown to the contrary, it is recommended that sites with
pH <6 be excluded from data sets prior to analysis.
In Case Studies I and II provided in Sections 4 and 5, sites with pH <6 were excluded
from the data set prior to analysis. Therefore, the case studies were developed without the
influence of pH, analogous to controlling for confounders in a laboratory test. Nevertheless,
field data show that even below pH 4.5, high SC was a stronger predictor than acidity on the
occurrence of Ephemeroptera (see Appendix B in U.S. EPA, 201 la). A contingency table
showed that Ephemeroptera were observed at low pH unless SC was high. Also, calculating the
HC05 using the data set from the EPA Benchmark Report (U.S. EPA, 201 la), with and without
the inclusion of low pH sites, yielded very similar results (295 [j,S/cm for all sites compared to
288 [j,S/cm pH <6 sites removed). Therefore, although EPA recommends the removal of pH
sites from the data set prior to analysis, there is evidence to suggest that the derived criteria are
applicable to all streams regardless of pH.
3.6.2.	Waterbody Type
Another important consideration when it comes to applicability of the field-based
approach is waterbody size and type. The EPA recommends analyzing the effect of catchment
size on the XCD model and documenting the decision, rationale, and supporting analyses for
3-32

-------
applicable water body types for SC criteria derived using this method. For example, in Case
Study I, the data used to derive the CCC are from perennial streams with catchments that range
from 0.34 to 17,985 km2. Literature reviews and analyses (described below) were performed in
the Problem Formulation phase of this assessment to determine relevant (applicable) waterbody
types. As a result, all stream types and sizes were included in the data sets for these case studies.
Although the field data used in the case studies were only collected from perennial
streams, available information from the open literature indicates that many of the
macroinvertebrate taxa persist in intermittent and perennial channels, albeit at different densities
and for varying amounts of time. For example, Grubbs (2010) assessed the relationship between
stream-flow permanence and macroinvertebrate community structure along temporary and
perennial hydrologic gradients in forested headwater streams in a Cumberland Plateau watershed
in the Kentucky River Basin. Grubbs found that the vast majority (91 out of 108) of
macroinvertebrate taxa were observed in both the perennial and temporary channels.
Macroinvertebrate taxa have many adaptations to survive temporary dry periods including egg
diapause, nymph aestivation, and nymph migration into hyporheic zones (the area beneath a
streambed, where shallow groundwater and surface water mix) or intermittent pools (Datry,
2012). Macroinvertebrates may use temporary stream resources for portions of their life cycle
(e.g., nursery habitat) and move downstream as they get older and larger and conditions require
emigration to areas of greater flow (De Jong and Canton, 2013; Feminella, 1996; Stout and
Wallace, 2003). These studies suggest that temporary streams are used, at least for a portion of
their life cycle, by many of the macroinvertebrate taxa considered in the XCD method.
Discharge to ephemeral streams ultimately affects downstream intermittent/perennial
streams (via gravity and flow through the tributary system during precipitation events). As a
result, addressing SC in upstream ephemeral streams is often critical to ensuring that downstream
aquatic life is not exposed to harmful levels of SC above the criteria.
Although intermittent and perennial streams are likely to have similar SC regimes, larger
catchments may not have the same background SC as smaller streams owing to hydrological
contributions from different geologies or other factors. Options include limiting the use of
derived criteria to the range of sampled catchments represented in the data set, developing
criteria for different stream classes, or demonstrating that there is no difference due to catchment
size.
3-33

-------
In the EPA Benchmark Report, larger streams (catchment areas >155 km2) were
originally screened from the data set that was used because sampling methods might differ for
nonwadeable streams (U.S. EPA, 201 la; Flotemersch et al., 2006). However, a subsequent
analysis indicated that 25 of the 30 most salt-intolerant genera based on derived XC95 values for
Ecoregions 69 and 70 (see Appendix E) were documented in these larger rivers (see also
Appendix B in U.S. EPA, 201 la). Inclusion of the data from large streams did not significantly
change the magnitude of the HC05 (289 (j,S/cm) compared to the HC05 without data from larger
systems (295 (j,S/cm). Additional analyses support that result. An analysis of 3,115 sites
(3,736 samples total: 1,661 in Ecoregion 69 and 2,075 in Ecoregion 70) with drainage areas up to
17,986 km2 suggests that SC and drainage area are very weakly correlated (r2 = 0.044, see
Figure 3-8). These are neither random samples nor reference streams and may not represent
natural background. The apparent background SC, estimated as the 25th centile of probability
sites, for streams draining areas >155 km2 in Ecoregion 69 and 70 are 148 and 188 [j,S/cm,
respectively; both of these estimates are within the confidence bounds for estimated background
SC using the example criterion-derivation data sets. Therefore, the example ecoregional criteria
in the case studies are relevant for all stream sizes. However, professional judgment is warranted
when applying the example criteria to streams crossing ecoregional boundaries and stream
catchments draining >1,000 km2 because they are less well represented in the data sets (see
Figure 3-8). For example, great rivers such as the Ohio and Mississippi Rivers were not
represented in the data set, and they cross many ecoregional boundaries.
3-34

-------
10000 -
E
o
«i
1000 -
"0
£=
o
o
o
Oj
Q_
CO
100 -
r = 0.25
Eco69
Eco70
m
•v »** * * + * *
' * **~ /* * ~ *
;. • ,~ .
" * •
~i—i i 111111
1
1111 n|
10
~1—I I I 11111—
100
i 111111	1—i i 11111|	r
1000	10000
Drainage area (krn^)
Figure 3-8. Correlation of specific conductivity and drainage area up to
17,986 km2, Spearman's r = 0.25. This analysis shows a very weak correlation
between specific conductivity and drainage area and supports inclusion of data
from all stream sizes in the data set for example criteria derivation. The fitted
lines are the locally weighted scatterplot smoothing (LOWESS, span = 0.75,
linear polynomial) for each data set.
3.7. METHODS FOR APPLICATIONS TO NEW AREAS
Not all areas of the country have sufficient water chemistry and biology data to derive
criteria for SC by the XCD method of calculating XC95 and HC05 values. For such cases, the
EPA is providing alternative methods that geographically extend results of the primary XCD
method (see Section 3.1). One alternative method extends criteria developed in one area to other
areas within the same ecoregion. This method is termed background matching. A second
alternative method estimates criteria for new areas with different background SC using a
background-to-criterion regression model. Both of these methods rely on the estimated
background SC.
The feasibility of applying a conductivity benchmark outside its area of derivation was
considered by the EPA SAB in their review of the EPA Benchmark Report (see Section 3.7
3-35

-------
Transferability of the Method to Other Regions in U.S. EPA, 201 lc). In general, the SAB
concluded that the numeric benchmark was applicable to the regions where the field data were
collected and could be applicable to other areas where sufficient data allow for evaluation of
applicability of the benchmark. Since then, the EPA has developed an understanding of the
mechanistic relationship between background SC and the extirpation of salt-intolerant
invertebrates. This relationship allows the EPA to relate HCos values from one area to another
based on background SC.
Background SC in a region and the associated HCos are expected to be strongly related
based on ecological and evolutionary theory and the observed responses of invertebrates to major
ions (see Section 2.4). The most salt-intolerant invertebrates occur in streams with the lowest
background SC (see Appendix D). As SC increases, the most salt-intolerant species are
adversely affected and ultimately cannot persist. As a result, where regional background SC is
higher, those taxa adapted to low SC are absent, and the SC level that is protective of 95% of
taxa (HCos) is higher.
The EPA developed the B-C method using 24 field XCDs from ecoregions with
background SC ranging from 22 to 626 [j,S/cm. Relatively salt-intolerant genera, as indicated by
low XC95 values, occupy habitats in each region with the lowest ionic concentration. When both
are log-scaled, the increase in background SC is linearly related to the HC05. This regular and
biologically relevant relationship between background SC and the HC05 confirms that the lower
portion of the XCDs are similar in similarly exposed communities even though the represented
genera may differ among ecoregions. The relationship between background SC and the HCos
identified from the XCD is sufficiently strong to identify a CCC for areas with sufficient stream
chemistry data but little or no paired biological data within an ecoregion or for new ecoregions
(see Appendix D).
The association between background SC and the HCos was used to develop the
background matching approach and the approach using the B-C method described in
Sections 3.7.1 and 3.7.2, respectively.
3.7.1. Application within an Ecoregion—Background Matching
If paired SC and biological data are not available for a new area within an ecoregion, the
background SC may be used to assess applicability of the derived CCC to the new area using a
3-36

-------
technique called background-matching. Regional background SC is defined here as the range of
SC naturally occurring in waters that have not been affected by human activity. Background SC
is important to water quality protection because it represents the ionic concentrations to which
organisms in the region are naturally adapted. Minimally affected waters with low SC play a
particularly important role by diluting polluted downstream waters and serving as refugia for
salt-intolerant organisms. The background-matching approach is demonstrated in Case Studies I
and II (see Sections 4 and 5) for the new areas within an ecoregion that were not included in the
original example criterion-derivation data sets.
In this discussion, the phrase, original area, refers to the geographic area from which the
data are obtained to develop SC criteria using the XCD method. The phrase, new area, refers to
a geographic area within the same Level III ecoregion that was not represented in the criterion
derivation data set. When applying field based SC criteria developed with data from the original
area to a new area, the background SC levels and the ion composition should be similar in both
areas. For instance, the example criteria are derived with data for streams where the ionic
mixture is dominated on a mass basis by ([SO42 ] + [HCO3 ]) > [CP],
The relationship between background SC and the HC05 identified from the XCDs is
sufficiently strong to identify a HC05 for areas without biological sampling within an ecoregion
or for new ecoregions. This B-C regression model was developed using biological data paired
with SC data from waters with ionic mixtures dominated by calcium, magnesium, sulfate and
bicarbonate ions and where background SC did not exceed 626 [j,S/cm. Therefore, the model is
most appropriate for waters with similar ionic characteristics. The model has not been
thoroughly tested and professional judgment is required for places where on a mass basis the
major ions are ([HCO3 ] + [SO42 ]) < [Cl~] or ([Ca2+] + [Mg2+]) < ([Na+] + [K+]). In particular,
the B-C model is not appropriate for waters dominated by NaCl (Haluszczak et al., 2013,
Entrekin et al., 2011; Gregory et al., 2011; Veil et al., 2004) or road salt (Forman and Alexander,
1998; Kelly et al., 2008; Environment Canada and Health Canada, 2001; Evans and Frick, 2001;
Kaushal et al., 2005). The B-C model may also be defensible for ionic mixtures dominated by
sodium, sulfate and bicarbonate ions (Brinck et al., 2008; Dahm et al., 2011; Jackson and Reddy,
2007; National Research Council, 2010; Clark et al., 2001; Veil et al., 2004). This is because the
toxicity of these mixtures are more similar to that of calcium, magnesium, sulfate and
3-37

-------
bicarbonate ions than the toxicity of NaCl (Mount et al., 2016; Kunz et al., 2013; Soucek and
Dickinson, 2015).
In the background-matching approach, the background for the ionic mixture of the new
area is compared with the background of the original area. If the 95% CI of the background SC
of the new area overlaps with the 95% CI of the background in the original area, the original
criterion is considered applicable. If the CIs for the two areas do not overlap, then a
dichotomous decision tree is used to guide further evaluations (see Figure 3-9). The
dichotomous decision tree for assessing the applicability of criteria from the original area to a
new area of an ecoregion may require a weight-of-evidence assessment described in detail in
Appendix C, calculation of an HCos using a regression model described in Section 3.7.2 and
Appendix D, or collection of sufficient data to derive a different HCos for the new area.
3-38

-------
Yes
No
No
Yes
Apply new
HCns
Derive a new HC05 by
the XC-Distribution or
B-C Method
Apply XC-distribution
derived HC05 to ecoregion
to the new area?
Using weight of evidence, is the
different background SC in the
new area naturally caused?
Is there overlap between the 95% CIs of the SC
background in the new area and in the area used to
derive SC HCm with the field XC-Distribution Method?
Figure 3-9. Method for selecting a criterion continuous concentration (CCC)
for a new area within an ecoregion using minimally affected background.
The hazardous concentration of the 5th centile of a taxonomic extirpation
concentration distribution (HCos) from a field derived extirpation concentration
distribution (field XCD) is one that has been previously developed using a large
data set. The background-to-criterion (B-C) method uses a regression model to
predict a criterion and confidence interval (CI) from background specific
conductivity (SC).
Portions of the same ecoregions in different political jurisdictions are expected to have
similar characteristics with respect to the primary factors that control background SC (Hem,
1985, Griffith, 2014, Olson and Hawkins, 2012, see Section 2.1). These primary factors are
underlying geology, physiography, and climate; secondary factors include soils and vegetative
cover (Olson and Hawkins, 2012; Griffith, 2014; Hem, 1985). Because Level III ecoregions
were delineated based on similar considerations (Omernik, 1987), the SC regime and ionic
composition of dissolved salts in streams within an ecoregion tend to be similar throughout.
However, there may be situations where it is not appropriate to apply criteria derived for the
ecoregion to a particular stream reach. For example, naturally lower or higher concentrations of
ions may occur due to subecoregional differences such as cross boundary influences, glacial
melt, salt springs, highly soluble rock, or other natural sources.
3-39

-------
3.7.1.1.	Obtaining a Data Set
The first step in the background-matching approach is to assemble the data sets from
sampled sites that are distributed across the full range of SC conditions in the new area. All else
being equal, the larger the data set, the more reliable the estimate of background SC. Next, sites
with qualitatively different ionic mixtures are removed from the data set. In the example case
studies, chloride-dominated sites are removed from the data set so that background SC is
estimated only for sites dominated by sulfate and bicarbonate (i.e., ([HCO3 ] + [SO42 ]) > [CI]
in mg/L). For three of the example case studies, Sections 4, 5, and 7, the dominant cations are
Ca2+ and Mg2+. For the example case study, Sections 6, the dominant cation is Na+.
3.7.1.2.	Estimating Background Specific Conductivity (SC)
If minimally affected background SC is not known, it can be estimated from field data
that are representative of SC throughout the year. In particular, the data set should not be biased
toward seasonal extremes by sampling only during seasons of freshets or droughts. Background
SC may be estimated as a proportion of a regional sample of sites or a sample of reference sites
that are judged to be among the best within a region (U.S. EPA, 201 la).
Regional samples from a random or probability-based design (Stevens and Olsen, 2004)
include all waters within the sampling frame, including impaired sites. To characterize the
minimally affected streams in a regional sample, the 25th centile is conventionally used
(U.S. EPA, 2000a). However, when land cover modification (or other anthropogenic
disturbance) is pervasive, selection of a centile lower than the 25th may be justifiable.
When estimating background concentrations using minimally affected reference sites, it
is conventional to use only the lower 75% of reference values (U.S. EPA, 2000a). One
indication of the need for a different centile is when reference sites have a broad range of SC
values suggesting that the reference condition contains some sites with anthropogenic
disturbance, or that the sites are not classified to partition natural variability (e.g., headwaters
draining through limestone glacial till into an area of weathered bedrock). An expanded list of
possible considerations is provided in Appendix C. When there is great confidence in the quality
of reference sites, a 90th centile may be used.
When there are sufficient good quality reference sites, the regional and reference methods
yield similar background estimates (NYSDEC, 2000, TDEC, 2000, U.S. EPA, 2000a). But, in
3-40

-------
general, estimation based on a random sample of the region tends to yield a more accurate
estimate of current background when there are sufficient data to characterize the full distribution
of SC in the region. Unlike the selection of reference sites, a random sample does not depend on
the original intent of data collection or the judgement of the data collectors (Whittier et al.,
2007a, b).
After estimating background SC for both the original and new areas using the method in
this section, the background-matching approach requires estimating their variability so that
confidence intervals can be calculated. The confidence interval for a background SC estimate
can be calculated using a bootstrapping technique. Bootstrapping is a statistical resampling
technique that is often used in environmental studies to estimate confidence limits of a
parameter. This bootstrapping application involves randomly resampling the original water
chemistry data set 1,000 times with replacement, storing the 1,000 data sets, calculating the
background for each data set, and then estimating the 95% CI for the mean of the set of
1,000 background values generated by the bootstrapping procedure. This is similar to the
procedure described in Section 3.1.3.1.
3.7.1.3. Background-Matching Approach
Once the means and confidence limits on the background SC in the original area and new
area have been estimated, they can be compared to determine whether they sufficiently match
using the decision criteria depicted in Figure 3-9.
1.	If the 95% CI of the background SC values from the new area overlaps with the 95% CI
of the background SC values from the original area, then apply the XCD derived HCos for
that ecoregion throughout the new area.
2.	If the 95%) CIs do not overlap, then use a weight-of-evidence approach to determine
whether the background in the new area represents natural or anthropogenic sources as
described in Appendix C.
3.	If the difference in 95%> CIs is due to anthropogenic alteration, then apply the XCD
derived HCos from the original area to the new area.
4.	If the difference in SC is naturally caused by geology, climate, or other natural factors,
then derive a new HCos for the new area using a sufficiently large and appropriate data
set from the new area (see Section 3.1) or calculate an HCos based on background SC for
the new area using the B-C regression method (see Section 3.8).
3-41

-------
For example, the decision matrix was used to determine the applicability of XCD SC
criteria from an original area to a new area using this background-matching approach (see
Example Case Studies in Sections 4 and 5).
3.7.1.4.	Options When Background in New Area is Different than in Original Area
If the estimated background SC is higher or lower in the new area and the ion
composition is similar compared to the original area, two possible causes should be considered.
First, differences may be due to natural geological factors (e.g., higher SC due to salt springs,
lower SC due to glacial melt, or other differences due to natural geological features) or to
climatological factors. In these situations, criteria for the new area can be developed using either
the XCD method or the B-C regression method. Second, differences may be due to widespread
anthropogenic changes that have increased the apparent background (e.g., due to irrigation,
agriculture, impervious surfaces, resource extraction, or acid deposition, etc.). In this second
situation, the criteria developed for the original area may or may not be applicable to the new
area.
To distinguish between these two possibilities, the cause of the apparent background can
be evaluated in a weight of evidence. Considerations may include analysis based on geology,
land use, ionic signatures and known inputs, historical and recent trend analysis, atmospheric
sources, discontinuities in background across political boundaries, identification of high and low
SC anomalies, stream size and connectivity, data set characteristics, sampling methods, and
biological evidence of past and present observation of susceptible genera (see Appendix C).
3.7.1.5.	Summary of Background Matching Method
The original criteria are applicable to a new area in the same region if the background SC
is not different from the background of the original area (i.e., overlapping 95% CIs) and the ionic
mixture is the same (e.g., in the case studies, ([HCO3 ] + [SO42 ]) > [CP] in mg/L). The original
criteria are not applicable to the new area if the background for the ionic mixture is different
owing to natural causes or if the ionic mixture is different. If the background SC is higher or
lower than the original background, a new criterion may need to be developed for the new area.
A weight-of-evidence analysis can be done to evaluate whether the difference in background SC
3-42

-------
is due to natural causes. When possible, an independent data set should be used to corroborate
the estimates of background SC.
3.7.2. Developing a Criterion Using Background-to-Criterion Regression Method
Because large sets of paired chemical and biological data are not available for all
ecoregions in the United States, the EPA developed a model to calculate a CCC using the
background SC of an ecoregion. The B-C regression method can be used in a new ecoregion or
a new area within an ecoregion with a different SC regime (e.g., at a scale smaller than Level III
ecoregions).
The relationship between minimally affected background SC and HCos for 24 Level III
ecoregions was characterized using least squares linear regression (see Figure 3-10). The
relationship between background SC and HCos was modeled and the association was strong
(r = 0.93), as was expected given the importance of the concentrations of major ions in defining
the tolerance of species (see Section 2.4). The relationship between background SC and HCos is
sufficiently reliable for identifying a CCC for areas without biological sampling within an
ecoregion or for new ecoregions (see Appendix D).
3-43

-------
1000 -
t-J
l/l
=L
>-
-t-j
'>
u
=3
"O
£Z
o
u
u
CJ
(U
Q.
LT)
Ln
o
U
500 -
100 -
50 "

0


0 '
J' 0
' ^ O
* "
0
y = 0.657*x + 1.075
i	1	1	1—i—i—i—i r
20	50	100
200
i	1	1	1—i—
500
25th Centile Specific Conductivity (|j.S/cm)
Figure 3-10. Empirical model of the 5th centile of a hazardous concentration
(HCos) and background specific conductivity (SC) estimated at the
25th centile for 23 distinct ecoregions (24 data sets). Solid line is the
loglO-loglO normal regression line; therefore, x and>' are loglO expressions.
Dotted lines demarcate the 50% prediction intervals, that is, the 50% probability
that any new HCos would plot within those bounds and only 25% below the lower
prediction limit (PL). The regression coefficient R2 = 0.87.
The B-C regression method shown in Figure 3-10 was derived using independent data
sets from 24 ecoregions (see Appendix D). First, SC XC95 values were estimated. From these,
24 genus-level XCDs were constructed and HCos values derived. Those HCos values were
regressed against the estimate of background SC for each ecoregion. In an ecoregion with low
background SC, very salt-intolerant taxa are represented. In an ecoregion with a moderate
background SC, taxa with an XC95 greater than the moderate background are likely to survive
and contribute to the XCD, whereas salt-intolerant taxa with XC95 values less than the moderate
background are not likely to contribute to the XCD. As XCDs are developed for ecoregions with
3-44

-------
increasingly higher background SC levels, each XCD begins at a higher background SC, and
thus the most salt-intolerant genera in an ecoregion occur at progressively higher SC levels. This
association is evidence that where low ionic concentration waters are present in an ecoregion,
organisms that are specialized for that niche are likely to inhabit them; and, where low ionic
concentration waters are not present, salt-intolerant species are not likely to occur. The resulting
B-C regression model provides a convenient method to predict an HCos from the minimally
affected or least disturbed background SC of an ecoregion. Descriptions of the derivation of the
regression model, the data sets used, and the individual XCD models are presented in
Appendix D.
The central tendency of a regression model is more robust than any single measurement.
For the purpose of model development, data requirements were relaxed relative to those for
calculating a HCos using the XCD method (i.e., fewer than 90 genera across 500 sites) (see
Appendix D for a description of data requirements for the B-C method). Individually, many of
the 24 HCos values used to develop the B-C method have not been subject to analyses needed for
development of a CCC and should be considered as provisional. For example, the HCos
estimates used in this model were not supported by full confounding analyses, as is described in
the EPA Benchmark Report (U.S. EPA, 201 la). However, the true HCos value is expected lie
between the upper and lower 50% prediction limit (PL). Values in Appendix D, Table D-3 are
provisional with a good degree of confidence owing to the larger sample sizes (>60 samples)
used to estimate background SC. Table D-4 lists ecoregions with background estimates based on
modest survey data sets (N= 20-60 samples) and would benefit from additional sampling to
confirm the calculated background SC and the calculated HCos. Table D-5 lists ecoregions
where the data set may represent fewer than 25% minimally affected streams and therefore are
protective of aquatic life in least disturbed streams. Table D-6 lists ecoregions that may not be
served by the B-C method because the ionic mixture is likely to be different (e.g., chloride
dominated), the estimated natural background SC exceeds the range of the model, and/or there
were fewer than 20 samples available. In all cases, the EPA recommends using the largest data
set possible to estimate background SC, understanding and accounting for areas with different
(higher or lower) background SC, and performing independent calculations to derive HCos
values.
3-45

-------
3.7.2.1. Using Background to Calculate a Hazardous Concentration of the 5th centile (HCos)
of a Taxonomic Extirpation Concentration Distribution (XCD)
The HCos for a defined geographic area or ecoregion without a sufficient data set or
without suitable biological data may be calculated using the B-C method based on the
background SC of that area or region. The decision tree for calculating a CCC from minimally
affected background is shown in Figure 3-11. Equations 3-1 and 3-4 can be used to calculate the
mean HCos and eq 3-5 calculates the lower 50% PL for the area or region. Sections 6 and 7
provide examples that use the decision tree to develop example criteria for 2 ecoregions in the
West, one with low and one moderately high background SC.
3-46

-------
NO'
YES
YES
NO
YES
NO
YES
YES
Apply mean
B-C modelec:
hc05
Apply XCD
derived HC
Apply lower
50% PL B-C
modeled HC05
Apply
new
hc05
Derive new HC05
by XCD method
>200 paired SC and
biological data?
XCD HC05 <
lower 50% PL
B-C modeled
hc05?
Calculate HC0S with
B-C model and
background SC
XCD HC05 >
mean
B-C modeled
HC05?
>500 paired SC and biological data
suitable for deriving SC HC05?
XCD HC0S < mean
modeled and > lower
50% PL modeled
HQ,,?
Background <626 p.S/cm?
[hco3] + [so4] > [CI]?
Derive HC05 by XCD method
and by the B-C model and
background SC
Figure 3-11. A decision tree for calculating and applying a hazardous
concentration of the 5th centile of a hazardous concentration (HCos).
This flow chart may be used when developing a criterion continuous
concentration (CCC) for new ecoregion, a new area within an ecoregion, or other
defined geographic area using the field extirpation concentration distribution
(XCD) method, background-to-criterion (B-C) method and minimally affected or
least disturbed background specific conductivity (SC). Numbered product paths
are described in the body of the text.
Where the background is less than 626 pS/cm and the waters have a similar ion
composition to those used to derive the model, the B-C method can be used (see Figure 3-11).
Where there are >200 but <500 sites with paired biological and SC data, HCos values are derived
using the XCD method and compared to the mean and lower 50% PL of the B-C model. These
3-47

-------
values are compared to select the appropriate HCos as follows. (1) If the XCD HCos (see eq 3-1)
is greater than the mean B-C modeled HCos (see eq 3-4), then the mean B-C modeled HCos is
recommended (see eq 3-4) as a conservative approach to account for uncertainty associated with
a smaller data set. (2) If the XCD HCos is between the mean B-C modeled HCos and the lower
50% PL, then the XCD HCos is recommended because the XCD from measured data from the
region is more likely to represent the region than the more general B-C model. (3) If the XCD
estimate is below the lower 50% PL, then the lower 50% PL is recommended as the HCos (see
eq 3-5). This is recommended because the XCD is calculated from a smaller data set. Also, it
may be overly protective because it is more uncertain than the modeled results which indicate
that 75%) of HCos values from areas with a similar background SC are estimated to be greater
than a value less than the lower PL. The lower 50% PL is also recommended when there are
fewer than 200 paired biological samples because there is no XCD for comparison. For both
situations, the SC data and the B-C model is used to estimate the HCos. (4) Where the
background SC is greater than 626 [j.S/cm, the range of the model is exceeded, and it is
recommended that data be collected to derive the HCos using the XCD method (see Section 3.1).
3.7.2.2. Formula for Calculating the Hazardous Concentration of the 5th centile of a
Taxonomic Extirpation Concentration Distribution (HCos) from the Background-to-
Criterion Model
The B-C model is described by the following formula:
7=0.657X+ 1.075.	(3-4)
Where:
Xis the loglO of the ecoregional background SC
Y is the loglO of the predicted HCos
3.7.2.3. Formula for Calculating the Lower and Upper 10% Prediction Limits
The upper and lower PL for a predicted loglO HCos value y can be calculated from the
regression line using eq 3-5 (Zaiontz, 2014) and loglO transformed SC values (x) as follows:
3-48

-------
9 ± Wln-2S J1 + J + = PL
(3-5)
Symbol	Explanation
y	Log 10 of the mean predicted HCos
n	Number of samples
a	Alpha error rate for prediction
interval (desired confidence level)
tn_2 Student's lvalue at specified
confidence level (alpha, a) and n-2
degrees of freedom
Sy	Residual standard error of
prediction (standard deviation)
SS	Sum of square of x deviation from
their mean, SS = £f=i(xi — x)2
x	Mean x values used in the model
generation
i°	A new loglO background (x) value
for a new prediction interval
PL Upper and lower prediction limits
of mean predicted y
Example from the 23 ecoregion B-C model
Variable differs for each case
n = 24
50% prediction interval (a = 0.5)
For 50% prediction interval (a = 0.5),
£(1—0.5)/2,24—2 = 0.686
Sy = 0.11
SS = 4.21
x = 2.15
SC value differs for each case
SC value differs for each case
The estimated backgrounds listed in Tables D-3, D-4, and D-5 of Appendix D for 62 of
85 Level III ecoregions were used to estimate the HCos from the B-C model. HCos values and
the lower 50% PLs were estimated using eqs 3-4 and 3-5 and estimated background from
probability survey data. Predicted base-flow SC (Olson and Hawkins, 2012) was used to assess
whether the 25th centile SC used in the calculation is minimally affected (see Tables D-3 and
D-4) or least disturbed background SC (see Table D-5). Although the B-C Model is strongly
log-linear within the sampled SC range, the EPA recommends estimation of HCos only for
ecoregions with a background <626 [j,S/cm to avoid extrapolation beyond modeled data. Some
regions may have different ionic matrices (e.g., chloride-dominant) for which the derivation of a
CCC using this method has not been verified. Those ecoregions are identified in Table D-6. The
decision tree depicted in Figure 3-11 was used to select example HCos values that if rounded to
two significant figures generates an example CCC.
3-49

-------
3.7.2.4.	Criterion Continuous Concentration (CCC) with <200 Paired Biological Data
When there are insufficient paired data, background SC is used to calculate the lower
50% PL of the mean HCos which is rounded to two significant figures to yield the CCC (see
eq 3-5). This result is shown in the Box 3 in Figure 3-11. The estimated CCC at the lower 50%
PL for 62 ecoregions can be found in Tables D-3, D-4, and D-5 in Appendix D.
3.7.2.5.	Criterion Continuous Concentration (CCC) with 200 to 500 Paired Biological Data
If a suitable paired biological and SC data set of 200 to 500 sites is available that meets
the requirements outlined in Appendix D.2.1, then the HCos is estimated from that data set using
the XCD method (see eq 3-1) and the B-C model (see eq 3-4). The lower of the two estimates is
recommended as the HCos unless the XCD estimate is below the lower 50% PL from the B-C
model (see Figure 3-11). This result is shown in Boxes 1 or 2 in Figure 3-11. If the XCD
estimate is less than the lower 50% PL of the HCos from the B-C model, then the lower PL is
used (see eq 3-5). This result is shown in Box 3 in Figure 3-11. In either case, the predicted
mean or the lower 50% PL HCos is rounded to two significant figures to yield the CCC. The
provisional or comparative values for the CCC based on the mean regression line for
62 ecoregions are shown in Tables D-3, D-4 and D-5.
3.7.2.6.	Calculation of the Criterion Maximum Exposure Concentration (CMEC)
A CMEC based on water chemistry data can be calculated as described in Section 3.2. If
there are insufficient data to calculate a CMEC, the upper 50% PL can be used to approximate a
CMEC.
3.7.2.7.	Summary
Although the B-C regression model is strong, there is scatter in the 24 HCos values, so the
lower 50% PL is used. In addition, when there are >200 and <500 paired biological and SC data,
the XCD method is applied to check the B-C model results. The B-C model can also be used to
evaluate estimates with data sets >500 when they do not meet other requirements for the SC
range of exposure, unbiased sampling, seasonal bias, etc. Section 6 provides an example case for
deriving an HCos for the Northwestern Great Plains, Ecoregion 43 in Montana, Wyoming, North
Dakota, South Dakota, and Nebraska. Section 7 provides an example case for deriving an HCos
3-50

-------
for the Level III Cascades, Ecoregion 4, in Washington, Oregon, and California. The estimation
of an HCos from background described here is a recommended approach for developing water
quality criteria for those ecoregions lacking sufficient data to develop one by the XCD method
from a regional data set.
3-51

-------
4. CASE STUDY I: EXAMPLE USING EXTIRPATION CONCENTRATION
DISTRIBUTION (XCD) METHOD IN A LOW BACKGROUND SPECIFIC
CONDUCTIVITY ECOREGION
This section presents a case study for the Central Appalachians (Ecoregion 69) to
illustrate how the analytical methods described in Section 3 can be used to derive example SC
criteria using the XCD method in an ecoregion with low background SC. Ecoregion 69 results,
including estimates of the CCC and CMEC, duration, frequency, and discussion of applicability
are included as examples to demonstrate the method. The derivations of the CCC and CMEC
analyses and results are based on data from Ecoregion 69 in West Virginia, and SC data from the
Ecoregion 69 outside of West Virginia was used to assess applicability of the criteria throughout
the ecoregion.
4.1. DATA SET CHARACTERISTICS
The Central Appalachians (Ecoregion 69) stretch from central Pennsylvania through
West Virginia and Kentucky to northern Tennessee with small portions in Maryland and
Virginia. The primary physiographic feature is a high, rugged plateau composed of sandstone,
shale, conglomerate, limestone outcroppings, and coal. Elevation ranges from 366 to 1,402 m,
with an average elevation of >790 m. Local relief between valleys and peaks can range from as
low as 15 m to as high as 594 m. Rainfall is highly variable due to the topographic diversity,
ranging from 96-152 cm/year, with the lowest rainfall in valleys and the highest at the peaks.
The region is characterized by distinct summer and winter seasons, with growing seasons in
agricultural regions (located within valleys) lasting as long as 165 days. However, pasture and
agriculture are limited owing to the rugged terrain, cool climate, and infertile soils. The
landcover is mostly forested with oak and northern hardwood forests. The high hills and low
mountains are covered by a mixed hardwood forest. Underground and surface bituminous coal
mines are common (Woods et al., 1999, 1996). Headwater streams in this ecoregion have some
of the freshest (lowest SC) water in the United States. These headwater streams play an
important role in diluting downstream waters that are anthropogenically impacted, and serve as
refugia for fish and other salt-intolerant organisms.
The data used in this case study are from a large field data set, the West Virginia
Department of Environmental Protection (WVDEP's) in-house Watershed Assessment Branch
4-1

-------
database (WABbase). Chemical and biological samples are from 1996-2011 and 1997-2010,
respectively. The WABbase contains data from Level III Ecoregions 66, 67, 69, and 70 in West
Virginia (U.S. EPA. 2010; Omernik, 1987; Woods et al., 1996). The WABbase data set provides
consistent sampling and analytical methods, high quality, broad spatial coverage and a large
number of perennial streams (2,299 distinct locations) in Ecoregion 69.
The WABbase contains data from a mixed sampling design that collects measurements
from long-term monitoring stations, targeted sites within watersheds on a rotating basin
schedule, randomly selected sample sites (Smithson, 2007), and sites chosen to further define
impaired stream segments in support of total maximum daily load (TMDL) development
(WVDEP, 2008a). Most sites are sampled once during an annual sampling period, but some
sites are sampled monthly for water quality. The data set contains water quality, habitat,
watershed characteristics, macroinvertebrate data (both raw data and calculated metrics), and
geographic location (WVDEP, 2008a). A wide range of SC levels were sampled, which is useful
for modeling the response of organisms to different levels of ionic concentration. The WABbase
includes assignation of reference status using a tiered approach. Analyses involving the use of
these reference sites were drawn from the Level 1 reference status (WVDEP, 2008b) which
selects reference sites that "are thought to represent the characteristics of stream reaches that are
minimally affected by human activities and are used to define attainable chemical, biological and
habitat conditions for a region" (WVDEP, 2013; Stoddard et al., 2006). Sites are initially
selected by a map coordinator based on GIS land use data that indicate minimal human
disturbance. Streamside, the appropriateness of the selected site is confirmed based on the level
of anthropogenic disturbance, lack of point discharges, habitat quality, pH, dissolved oxygen,
and SC (>500 (j,S/cm) is used to flag a site for further investigation before inclusion as a
reference site (WVDEP, 2013).
Macroinvertebrate records in the data set are based on collections from a total of 1 m2
area of a 100 m reach at each site. Using a 0.5 m wide rectangular kicknet (595 [j,m mesh), four
0.25 m2 riffle areas were sampled. In narrow or shallow water, nine areas were sampled with a
0.33 m wide D-frame dipnet of the same mesh size. Composited samples were preserved in 95%
denatured ethanol. A random subsample of 200 individuals (+20%) was identified in the
laboratory. All contracted analyses for chemistry and macroinvertebrate identification followed
WVDEP's internal quality control and quality-assurance protocols (WVDEP, 2008b, 2006).
4-2

-------
Quality assurance of the data set was judged by the EPA to be excellent, based on the database
itself and supporting documentation.
Several data filters, described in Section 3.1 (see Figure 3-2), were applied prior to
finalization of the data set and analyses. A total of 9,806 records from Ecoregion 69 are included
in the data set; of these, SC measurements were included in 8,989 samples. Many of these are
measurements of water quality without biological sampling. There are 1,911 paired samples
with SC measurements and biological samples. Of these, a total of 250 samples were removed
from the data set due to low pH <6 (237 samples) and high proportion of chloride ions
([HCO3 ] + [SO42 ]) < [CI-] (13 samples). Additional criteria were applied to identify
macroinvertebrates for inclusion in the example extirpation concentration distribution:
occurrence at reference sites and occurrence in 25 or more samples. Of the
219 macroinvertebrate genera identified in this ecoregion in the WABbase, 193 genera occurred
at least once at one of the 64 identified reference sites where invertebrate samples were collected.
A total of 142 genera occurred at 25 or more sampling locations. The final example
"Criterion-data set" has 1,661 samples belonging to 1,420 sites (stations) (depicted in
Figure 4-1). Of these 1,661 samples, 186 (11.2%) were sampled more than once between 1996
and 2010. Summary statistics for the data set used to derive the example CCC is shown in
Table 4-1. The statistical package R, Version 2.12.1 (December 2010), was used for all
statistical analyses (R Development Core Team, 2011).
SC ranged from 15.4-3,794 [j,S/cm which allowed the response of organisms to be
modeled for a wide range of SC levels. Scatter plots of parameters and SC are depicted in
Appendix A. 1.
4-3

-------
Pennsylvania
Ohio
West Virginia
Virginia
>a*&'••• ::
Kentucky
XV. vVk _ •• ••
*!»» • <5>'V ;«*•{ * ; .
7*
Legend
• Sampling stations
Ecoregion 69
120
160
kilometers
Figure 4-1. Ecoregion 69 extends from central Pennsylvania to northern
Tennessee. Sampling sites (stations) (N= 1,420) in the example Criterion-data set
that were used to derive the criterion continuous concentration (CCC) are
indicated as points.
4-4

-------
Table 4-1. Summary statistics of the measured water-quality parameters used to
derive the example specific conductivity criteria in Ecoregion 69.
The example Criterion-data set has 1,661 samples belonging to 1,420 stations.
Parameter
Units
Min
25th
50th
75th
Max
Geomean
N
SC
(iS/cm
15.4
94
229
540
3,794
225
1,661
Hardness
mg/L
2.18
28.03
64.31
132.7
1,492
64.43
834
Total alkalinity
mg/L
2
14
41
90
560
37
1,144
SO,2
mg/L
1
12
32
126
2,097
39
1,146
Chloride
mg/L
0.5
2
3
8
650
4
930
SO,2 + HCO,
mg/L
8.66
36.3
99.4
252
2,256
99.3
1,142
Ca, total
mg/L
0.67
6.9
16.9
33.5
430
15.8
842
Mg, total
mg/L
0.5
2.4
5.0
12
204
5.6
832
Na, total
mg/L
0.5
1.8
3.5
13
423
5.2
166
K, total
mg/L
0.5
0.7
1.2
2.4
16
1.4
164
TSS
mg/L
1
3
3
5
80
4
1,151
Fe, total
mg/L
0.02
0.09
0.18
0.38
4.9
0.19
1,170
Fe, dissolved
mg/L
0.01
0.02
0.03
0.07
1.1
0.04
995
Al, total
mg/L
0.01
0.06
0.1
0.19
3.3
0.11
1,142
Al, dissolved
mg/L
0.01
0.02
0.03
0.06
0.9
0.04
1,007
Mn, total
mg/L
0.003
0.02
0.03
0.06
4.4
0.03
1,142
Se, total
mg/L
0
0.001
0.001
0.003
1.3
0.002
665
DO
mg/L
2.06
8.47
9.27
10.2
17.1
9.41
1,644
Total phosphorus
mg/L
0.01
0.02
0.02
0.02
1.3
0.02
897
NOx
mg/L
0.01
0.14
0.28
0.45
11
0.26
910
Fecal
Counts/100 mL
0.5
15
65
300
250,000
71
1,405
pH
SU
6.01
7.00
7.54
7.97
10.48
7.48
1,661
Catchment area
km2
0.34
4.36
17.6
65.2
17,986
19.3
1,408
Temperature
°C
-0.28
14.2
17.9
20.7
30.2
17.5
1,661
RBP lOSc
RBP score
53
126
142
156
195
140
1,641
RBP 7Sc
RBP score
30
84
98
110
137
97
1,647
Embeddedness
RBP score
1
11
13
16
20
13
1,649
Percentage fines
(sand + silt)
-
0
10
12
20
100
15
1,620
All means are geometric means except pH, DO, Temperature, and Habitat Scores.
RBP = rapid bioassessment protocol (Barbour et al., 1999; RBP lOSc has 10 parameters while RBP 7 does not
include 3 flow-related parameters); TSS = total suspended solids.
4-5

-------
4.1.1. Background Specific Conductivity
Background SC was estimated at the 25th centile of the subset of probability-based
samples from the example Criterion-data set because its sampling design more closely matched
the ecoregional EPA-survey data set. Using this probability-based subsample of the WABbase
data set, the estimated background was 80 [j,S/cm (25th centile, 585 samples from 544 sites; see
Figure 4-2). Background was also estimated to be 63 [j,S/cm based on field data from reference
sites from the WABbase data set (75th centile, 112 samples from 82 reference sites; see
Figure 4-3). By comparison, the 25th centile was 94 [j,S/cm for all samples (reference and
nonreference sites) from the example Criterion-data set that was used to derive the HCos
(1,661 samples from 1,420 sites; see Figure 4-4). The monthly 25th centiles of
probability-sampled sites (see Figure 4-2) and all samples in the data set (reference and
nonreference sites, see Figure 4-4) were relatively consistent and at or below 100 [j,S/cm except
in July through October (see Figure 4-2). The effects of seasonal variability of SC on the
subsequent analyses was further evaluated and are presented in Appendix A. The large size of
the data set and the wide range in SC levels in the example Criterion-data set allowed for
characterization of the XC95.
4-6

-------
tooo -
1
§ 500 -
S"
-! 200 -
e
o
o
si 100 -
o
gj
o.
to
50 -
20 -
Figure 4-2. Box plot showing seasonal variation of specific conductivity
(jiS/cm) from probability-sampled sites from Watershed Assessment Branch
database (WABbase) 1997-2010. This represents a total of 544 sites with
585 samples from 1997-2010 from Ecoregion 69 with pH >6. Note the
difference in scale along the j'-axis between Figure 4-2 (probability-sampled sites)
and Figure 4-3 (reference sites). There are only eight October samples. See
Table 4-2 for number of samples per month.
Apr
May
Jun
Aug
Sep
Oct
Month
4-7

-------
200 -
100 -
a>
a.
m
i	1	1	1	1	1	1	1	r
Jan Feb Mdi Apr rid Jun Jul Aug -Sep Dec
Month
Figure 4-3. Box plot showing seasonal variation of specific conductivity
(jiS/cm) in the reference streams from Watershed Assessment Branch
database (WABbase) 1997-2010.
A total of 112 samples from 82 reference stations were used for this analysis to
estimate background specific conductivity. Please note the smaller scale on the
j'-axis compared to Figures 4-2 and 4-4. A total of 112 samples from 82 reference
stations were used for this analysis to estimate background specific conductivity.
Please note the smaller scale on the _y-axis compared to Figures 4-2 and 4-4. See
Table 4-2 for number of samples per month.
4-8

-------
2000 -
500
200
® 100
Q.
m
50
20
I I I I I I I I I I I
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Dec
Month
Figure 4-4. Box plot showing seasonal variation of specific conductivity
(jiS/cm) from all Ecoregion 69 sites from Watershed Assessment Branch
database (WABbase) 1997-2010 used to develop the example criteria.
This represents a total of 1,661 samples from 1,420 sites from 1997-2010. Note
the difference in scale along the j'-axis between Figure 4-4 (all sites, reference and
nonreference) and Figure 4-3 (reference sites). See Table 4-2 for number of
samples per month.
4.1.2. Ionic Composition
The ionic composition of the samples in the Ecoregion 69 data set was assessed to ensure
that the example criteria were derived for waters dominated by sulfate and bicarbonate anions
(see Figure 4-5). Of the 1,674 samples after low pH samples were removed, 56% of samples
(938 in total) included measures of calcium, magnesium, sulfate, bicarbonate, and chloride. All
but 13 sites (>98%) were dominated by bicarbonate and sulfate anions
([FICO3 ] + [SO42 ]) > [CI-]. The 13 chloride-dominated sites were excluded from the derivation
4-9

-------
analysis but are shown in Figure 3. Sodium and potassium were less frequently measured, but
did not exceed calcium and magnesium where measured for samples in the data set. Sites with
no ion measurements were retained in the data set because the data had shown that >98% of
samples were dominated by bicarbonate and sulfate anions; thus, it is expected that less than 2%
of samples in the Ecoregion 69 Criterion-data set are chloride-dominated.
The analysis may also be defensible for mixtures dominated by sodium, sulfate and
bicarbonate ions, e.g., produced water from deep coal mines (Thomas, 2002; Mayhugh and
Ziemkiewicz, 2005). This is because the toxicity of these mixtures are more similar to that of
calcium, magnesium, sulfate and bicarbonate ions than the toxicity of NaCl (Mount et al., 2016;
Kunz et al., 2013; Soucek and Dickinson, 2015).
4-10

-------
1000
CD
E
o
10
100
1000
S0~ 4- HCOj (mg/L)
Figure 4-5. Scatter plot of relationship between [CI ] and
([HCOs] + [SO42 ]) concentrations in streams in Ecoregion 69 data set from
1997-2010 with ionic measurements. Most (98.6%) of the samples (n = 938) are
below the diagonal line representing the separation between
([FICO3 ] + [S042~])-dominated and CP-dominated mixtures. Sites above the 1:1
line were excluded from the example Criterion derivation data set. Samples
depicted here include all sites regardless of pH.
4.1.3. Seasonal Specific Conductivity Regime
For this case study, chemical, physical, and/or biological samples were collected during
the sampling years 1997-2010 (January-December). Most (>85%) sites were sampled once
during an annual sampling period, but some (e.g., sites being studied to improve stream
condition within the TMDL Program) were sampled monthly for water quality parameters (see
Table 4-2).
4-11

-------
Table 4-2. Number of samples with reported genera and specific
conductivity meeting acceptance criteria for the Ecoregion 69 analysis
Number of
samples8
Month
Total
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Full data set
8
4
36
159
269
163
197
388
342
89
0
6
1,661
Probability
sites
0
0
0
66
190
116
84
68
53
8
0
0
585
Reference
sites
6
3
3
10
19
12
27
24
4
0
0
4
82
Percentage
of total
0.5
0.2
2.2
9.6
16.2
9.8
11.9
23.4
20.6
5.4
0
0.4
(100)
aNumber of samples is presented for each month.
Samples collected from the WVDEP-identified reference sites indicate that SC levels are
generally low and similar throughout the year, although slightly higher in September (see
Figure 4-3). These data show that SC concentrations in flowing waters in the study area can vary
somewhat by season, likely depending on stream discharge, rainfall, snowmelt, and other
hydrological factors. As described in Section 3.1.4 (and in greater detail in the EPA Benchmark
Report), the effects of seasonal differences in SC levels and aquatic insect life history were
evaluated by comparing HCos values partitioned for season. After careful consideration of the
similarity between the spring HCos and the HCos based on the full data set at the low end of the
XCD, the example ecoregional criteria were derived using all available data, regardless of the
time of year they were collected (see Sections 3.1.4 and Appendix A.2 in this assessment, and
U.S. EPA, 2011a).
4.2. RESULTS
4.2.1. Extirpation Concentration (XC95) and Hazardous Concentration (HC05) Values
(Example Criterion Continuous Concentration [CCC])
The Ecoregion 69 example Criterion-data set (see Table 4-1) was used to develop XC95
values from weighted CDFs. The histogram used to develop weights is depicted in Figure 4-6.
The XC95 values that were used in the XCDs are listed in the order of least to most salt-tolerant
in Appendix A.3. The generalized additive model plots used to designate ~ and > values for
those XC95 values are depicted in Appendix A.4. The weighted CDFs used to derive the XC95
4-12

-------
values are shown in Appendix A.5. The HCos for Ecoregion 69 was calculated at 305.4 [j,S/cm
(see Figures 4-7 and 4-8); the two-tailed 95% confidence bounds were 233-329 [j,S/cm. Those
bounds, derived by bootstrap resampling, indicate that different data sets could yield HCos values
within that interval. Rounding to two significant figures, the example CCC for Ecoregion 69 is
310 [j,S/cm.
O
to
O
e
O
•a*
o-
05
O
CM
O
dH
T—I I I I I I
100
1 HJU
Specific conductivity (|jS/cm}
Figure 4-6. Histograms of the frequencies of observed specific conductivity
values in samples from Ecoregion 69 sampled between 1997 and 2010. Bins
are each 0.017 (1/60) of the range of loglO specific conductivity units wide.
4-13

-------
GD
O
CD
O
CN
O
o
o
30 b uS/cm
r
200
500
luuu
JilOll
Specific conductivity (pS/cm)
Figure 4-7. Example genus extirpation concentration distribution (XCD) for
Ecoregion 69. Each point is an extirpation concentration (XC95) value for a
genus. There are 142 genera. The hazardous concentration (HC05) is 305 [j,S/cm
(95% confidence interval is 233-329 (j,S/cm) and is the specific conductivity at
the intersection of the XCD with the horizontal line at the 5111 centile. XC95 with
an approximate or greater than designation are shown as triangles.
4-14

-------
o
FO —
o
CM
T3
m
tu
Q.
m
w
m
OD
£Z
O
=E
o
o.
o
o
04
o
305 pS/cm	Jt
J
HI
o


Q
o _J
100
I	I	I
200	500	1000
Specific conductivity Qj Start)
Figure 4-8. The lower end of the example genus extirpation concentration
distribution for Ecoregion 69. The dotted horizontal line is the 5th centile. The
vertical arrow indicates the hazardous concentration (HCos) of 305 [j,S/cm (95%
confidence intervals 233-329 (j,S/cm). Only the 50 most salt-intolerant genera are
shown to better discriminate the points on the left side of the distribution. The six
most salt-intolerant genera (i.e., extirpation concentration [XC95] < 305 (j,S/cm)
are Leptophlebia, Remenus, Pycnopsyche, Paraleptophlebia, Bezzia, and
Alloperla). XC95 values with an approximate or greater than designation are
shown as triangles.
4.2.2. Example Criterion Maximum Exposure Concentration
At sites meeting the CCC of 310 [j,S/cm, 90% of the SC observations are estimated to
occur below the CMEC (see Section 3.2). The CMEC was derived using the full Ecoregion 69
data set (9,806 samples collected between 1996-2011). Of the 9,806 samples in this ecoregion,
there are 5,823 samples in a July to June rotating year representing 564 rotation years,
536 unique stations, with at least 1 sample from July to October and 1 sample from March to
4-15

-------
June, and at least 6 samples within a rotation year (see Table 4-3). Note that inclusion of
samples is not contingent on biological data. Reference and nonreference sites were included to
ensure a range of SC (see Table 4-1).
Table 4-3. Summary data related to the calculation of the example criterion
maximum exposure concentration (CMEC) for Ecoregion 69
Number of samples July to June prior to biological sampling
5,811
Number of unique stations/rotation years
536/564
Number of WVDEP reference sites
15
ccc
310 (j,S/cm
CMEC
630 (j,S/cm
Of the 564 rotation years (536 unique stations) with multiple SC measurements, the
variability of within station SC was found to differ for streams with different mean SC (see
Figure 4-9). The locally weighted scatterplot smoothing (LOWESS) lines indicated that the
average variability (residual standard deviation for a station) in the middle of the SC gradient is
slightly higher than both the lower and higher ends of the gradient. The stations with annual
mean SC values between the 25th and 75th centile (which is approximately between
120-520 (j,S/cm) have relatively similar variances, and therefore, could be used to estimate the
standard deviation components of annual mean SC (310 (j,S/cm). There are 2,855 samples from
278 station years (265 stations) in the selected data sets for Ecoregion 69 with streams having
mean SC values between 120 and 520 [j.S/cm. The grand mean and standard deviation of this
data set were determined and the CMEC was calculated. The example CMEC calculation is
shown below:
CMEC for Ecoregion 69: I0log10(310) + 1-28*0.243 = 634 j ^s/cm
The example CMEC (see Table 4-3) rounded to two significant figures yields a CMEC of
630 [j.S/cm for Ecoregion 69. At this level, where the annual average SC is <310 [j,S/cm, 90% of
the observations are expected to be less than the CMEC.
4-16

-------
50 100	500 1000
Station mean specific conductivity
Figure 4-9. Illustration of within site variability (residual standard deviation
for each station) along the specific conductivity gradient (station mean) in
Ecoregion 69. The x-axis is log annual mean specific conductivity. Each dot
represents a station. The fitted line is a locally weighted linear scatterplot
smoothing (LOWESS, span = 0.75, linear polynomial model), while the two
vertical dashed lines represent logarithm mean specific conductivity of 120 and
520 [j,S/cm, respectively. Within those bounds the standard deviation is fairly
constant.
4.3. GEOGRAPHIC APPLICABILITY
The geographical applicability of the criteria throughout Ecoregion 69 was assessed using
the background-matching approach (see Section 3.7.1). The background SC of the new area
(i.e., the portion of Ecoregion 69 beyond the original data set) was estimated at the 25th centile
4-17

-------
(see Section 3.7.1.2; and Cormier and Suter, 2013a) and compared with the background SC
estimates for the original data set.
Because the example SC criteria presented here have been developed for a dissolved
mixture dominated by sulfate and bicarbonate anions, ([HCO3 ] + [SO42 ]) > [CP] in mg/L, all
chloride-dominated samples, ([HCO3 ] + [SO42 ]) < [Cl~] in mg/L, were removed from the data
set before estimating background SC. Thereby, the background for the new area is estimated for
the same ionic mixture as the example criteria.
4.3.1. Data Sources
Two data sets were used for this example applicability assessment: the original data set
used to derive the HC05 described in Section 4.1 and an EPA-survey data set (see Table 4-4).
The EPA-survey data set was used to evaluate and characterize ion concentrations and
water chemistry in the ecoregion (see Table 4-5). The primary sources of the combined data are
from EPA survey programs including the National Rivers and Streams Assessment (NRSA)
2008-2009 surveys (U.S. EPA, 2013b, 2009), Wadeable Streams Assessment (WSA) 2004
survey (U.S. EPA, 2006), Environmental Monitoring and Assessment Program (EMAP)
1993-1998 and Regional-EMAP (R-EMAP) 1999 surveys (U.S. EPA, 2013c), and National
Acid Precipitation Assessment Program (NAPAP) 1986 survey (NADP, 2013). Data sets are
based on random samples from June through September. Most report SC, alkalinity, hardness,
sulfate, chloride, bicarbonate, pH, and other water quality parameters. Ecoregions and sampling
sites are shown in Figure 4-10. All samples were collected from first-through fourth-order
streams as part of a probability-based design intended to estimate proportions of parameters for
various stream classes. The probability-design weights were not used in this characterization.
Analysis of water chemistry samples followed EPA procedural and QA/QC protocols from
EMAP (U.S. EPA. 2001, 1998b, 1994, 1987), Wadeable Streams Assessment (U.S. EPA, 2004a,
b), the NRSA (U.S. EPA, 2009), and NAPAP (Drouse et al., 1986; U.S. EPA, 1987). These data
sets were also selected so that methods would be comparable across the data set, and because
these studies used probability-based designs (i.e., randomly assigned sampling locations).
4-18

-------
Table 4-4. Description of survey data sets combined to form the EPA-survey
data set used to assess applicability of the example ecoregional criteria
throughout Ecoregion 69
Data set
Sampling period
Total N
KY
MD
PA
TN
VA
MAHA EMAP
1993-1995
42
0
3
35
0
4
MAIA EMAP
1997-1998
12
0
0
8
0
4
WSA
2004
9
3
1
0
3
2
NRSA
2008-2009
8
4
0
1
2
1
NAPAP
1986
41
2
6
29
4
0
Region 4 Wadeable
Streams R-EMAP
1999-2002
9
7
0
0
2
0
Total
121
16
10
73
11
11
MAHA = Mid-Atlantic Highland Assessment; MAIA = Mid-Atlantic Integrated Assessment.
4-19

-------
Pennsylvania
>!•
IVi
MD
Ohio
West Virginia
Virginia
Original
area
Kentucky
Legend
EPA combined data (Eco69)
Ecoregion 69
120 180
Kilometers
Tennessee
240
Figure 4-10. Ecoregion 69 extends from central Pennsylvania to northern
Tennessee. Sampling sites in the EPA-survey data set that were used to estimate
background in the "new" area for Ecoregion 69 are indicated as points. The
figure depicts 121 samples from 121 stations.
4.3.2. Geographic Applicability Results
A summary of water quality for the EPA-survey data set (see Section 4.3.1) for
Ecoregion 69, including major ionic constituents, is provided in Table 4-5. Background SC in
the new area was estimated from the full EPA-survey data set because no sample was dominated
by chloride ions.
Background SC for bicarbonate and sulfate dominated waters estimated as the 25th centile
of the EPA-survey data set for the new area in Ecoregion 69 (outside the area used to develop the
example criteria) was 63.5 [j,S/cm (95% CI 46-89 (j,S/cm) (see Table 4-6). The 25th centile from
the probability sample from the example Criterion-data set was 66 [j,S/cm (95% CI 60-75) (see
Table 4-6). The confidence bounds for background estimated from the example Criterion-data
4-20

-------
set overlap with the confidence bounds for background estimated for the rest of Ecoregion 69.
Therefore, the background SC regime throughout Ecoregion 69 appears to be similar, and thus,
the example criteria (CCC = 310 [j,S/cm, CMEC = 630 (j,S/cm) are considered geographically
applicable throughout the ecoregion. Other estimates of background from the reference sites in
the example Criterion-data set (63 [j,S/cm; 95% CI 60-65 (j,S/cm) and the example Criterion data
set (94 [j,S/cm; 95% CI 86-101 (j.S/cm) also overlap with the CI of the background for the rest of
Ecoregion 69 (see Table 4-6).
Table 4-5. Summary of water quality parameters for Ecoregion 69 from the
EPA-survey data set excluding the sites in West Virginia
Ecoregion
Ion
Min
Centile
Max
Relevant
N
10th
25th
50th
75th
Ecoregion 69
HCO3 (mg/L)
0.0
0.1
1.3
12.5
37.3
241.8
102
S042 (mg/L)
3.2
7.6
10.0
21.4
136.3
1,622.8
112
cr (mg/L)
0.5
1.1
1.7
3.0
8.6
59.0
112
Ca2+ (mg/L)
1.2
2.2
5.6
13.7
39.1
186.0
112
Mg2+ (mg/L)
0.6
1.0
1.5
4.5
17.3
152.1
112
Na+ (mg/L)
0.2
0.6
1.1
2.9
9.3
93.4
112
K+ (mg/L)
0.4
0.6
0.8
1.3
2.2
8.0
112
pH (SU)
3.0
4.7
6.2
7.1
7.7
8.6
121
3(hco3 + so42 )/cr
1.4
3.0
5.4
13.6
41.4
497.5
102
SC (|iS/cm)
23.7
34.5
63.5
183.5
426.8
2,515
121
aValue within category calculated from individual sample ion concentrations. HCO3 + SO42 /CI in mg/L greater
than 1 indicates the mixture is dominated by HCOs, + SO r .
4-21

-------
Table 4-6. Background specific conductivity estimates for Ecoregion 69
Data set
Centile used to
estimate
background
Background
jiS/cm
Confidence
interval
jiS/cm
Relevant N
(sites/samples)
EPA-survey data set from geographic
area not represented in the example
criterion derivation data set
25th
64
46-89
121/121
WABbase data set, probability sample
subset
25th
66
60-75
544/583
WABbase data set, reference sample
subset
75th
63
60-65
82/112
Example criterion derivation data set,
full data set
25th
94
86-101
1,420/1,661
4.4. SUMMARY OF EXAMPLE CRITERIA FOR ECOREGION 69
The case example for Ecoregion 69 includes an annual geometric mean (i.e., CCC) and a
1-day mean (i.e., CMEC), not to be exceeded more than once in 3 years on average. Both of
these distinct expressions of the example SC criteria would need to be met in order to adequately
protect aquatic life. These values indicate that freshwater animals in Ecoregion 69 would be
protected if the annual geometric mean SC concentration in flowing waters does not exceed
310 |aS/cm and the 1-day mean does not exceed 630 |aS/cm, more than once every 3 years on
average. These example criteria would apply to all flowing freshwaters (ephemeral, intermittent,
and perennial streams) in Ecoregion 69 inclusive of portions of Kentucky, West Virginia,
Maryland, Virginia, Tennessee, and Pennsylvania. On a site-by-site basis, these example
ecoregional criteria apply if the ionic mixture is dominated by anions of bicarbonate and sulfate.
For streams crossing into Ecoregion 69, professional judgment may be needed to assess the
potential effect of different ionic composition or concentration. Professional judgment is
recommended when applying to sites with a catchment area greater than 1,000 km2 (386 mi2)
owing to lesser representation in the example data set by this class of stream. On a site-by-site
basis, alternative specific conductivity criteria may be more appropriate if the natural
background of a site is shown to be lower or higher than its regional background specific
conductivity.
4-22

-------
4.5. EXAMPLE CRITERION CHARACTERIZATION
4.5.1. Factors Potentially Affecting the Extirpation Concentration Distribution (XCD)
Model
An assessment of potential confounders and an analysis of the influence of habitat quality
and sampling date for Ecoregion 69 can be found in Appendix A.2.
4.5.1.1. Sensitivity Analyses
As the minimum number of occurrences of a genus for inclusion in the data set increases,
fewer genera are included in the XCD. The HCos increases greatly when a taxon in the lower
5th centile is removed because it does not meet the minimum number of samples and then more
slowly alternates between increasing and decreasing as genera either above or below the
5th centile are removed because they do not meet the minimum number of samples (see
Figure 4-11). The pattern repeats until all genera above and below the lower 5th centile have the
same XC95 value (not shown). To maximize the number of genera included in the XCD, a
minimum of 25 occurrences was utilized.
4-23

-------
o
CO
320
300
&
:> 280
o
=3
"O
a
a
o
o_

-------
800 -
700 -
E
u
3. 600
CL)
— 500
CD
>
un
<-> 400
300 -
200 -I
+

500
1000	1500
Sample Size
2000
- 152
O
(J
130 X
- 109
87
- 66
- 44
23
ro
aj
c
O)
(3
aj
_Q
E
13
Figure 4-12. The effect of the size of the data set used to model the
hazardous concentration (HCos) based on the Ecoregion 69 example
Criterion-data set. As size of the data set increases, the number of genera
included in the extirpation concentration distribution (XCD) increases (triangles).
The HC05 stabilizes reaching an asymptote at approximately 500-800 sites
sampled (circles) and 90-120 evaluated genera.
4.5.2. Validation of the Extirpation Concentration Distribution (XCD) Model
The XCD model was validated and uncertainty around the HCos values was estimated
using bootstrapping, as recommended by the EPA SAB in their review of the EPA Benchmark
Report (U.S. EPA, 201 lc). The median HCos estimated from bootstrapping was 281 |iS/cm
(95% CI 233-329 (iS/cm) which is similar to the HCos of 305 }iS/cm measured using a 2-point
interpolation of the original XCD. The similarity between the two HCos values indicates a
similar model would be generated using an independent data set (see Figure 4-13).
4-25

-------
in
CM
TJ
0
Qj
CM
•*—>
03
0
Cl

1	



X

CD
kO
ra
T	


CD
0
cz

CD

O

4—

O
0
C
•*—
O
0
iz:

0

Cl

0
I—
kO
Q_
0
o
o
o
/ ^
/ •" '
f	•- /
( rf
'	• ¦¦'
/
'	+f
v ¦
/ •
/V
/ • '
	'?	¦*"!*	
*/
. •/
200
-i	|	i i i i [—
500	1000
Specific conductivity (pS/cm)
Figure 4-13. Cumulative distribution of the extirpation concentration (XC95)
values for the 25% most salt-intolerant genera (blue circles) and
95% confidence intervals (dotted lines) based on 1,000 extirpation
concentration distribution (XCD) bootstrapping results. Each tiny gray dot
represents an XC95 value for a bootstrapping iteration (note that the genera in each
percentage varies with each XCD iteration). Each larger blue filled dot represents
the calculated XC95 of the XCD for the criterion continuous concentration (CCC).
The median bootstrapped hazardous concentration (HC05) is 281 (iS/cm.
4.5.3. Duration and Frequency
Numeric criteria include magnitude (i.e., how much), duration (i.e., how long), and
frequency (i.e., how often) components. Appropriate duration and frequency components of
criteria are determined based on consideration of available data and understanding the
exposure-response relationship in the context of protecting the aquatic life of a water body. The
significant consideration used in setting the duration component of aquatic life criteria is how
long the exposure concentration can be above the criteria without affecting the endpoint on
which the criteria are based (U.S. EPA, 1991, 1985). Based on the temporal resolution of the
available field data set and an analysis of within-site variability of SC levels, EPA developed two
4-26

-------
different expressions for the example SC criteria in order to provide adequate protection for
aquatic life.
In this case, the majority (>85%) of sites used to derive the example CCC for
Ecoregion 69 were sampled once during an annual sampling period and thus represent the
average stream SC and macroinvertebrate assemblage information over the course of 1 year. As
a result, the appropriate duration for the CCC is 1 year. The duration for the CMEC, a level of
protection from acutely toxic exposures, is 1 day based on a review of the literature regarding the
onset of macroinvertebrate drift in response to elevated SC (see Section 3.3). At sites meeting
the CCC, 90% of the SC observations are estimated to occur below the CMEC.
EPA anticipates that an appropriate allowable frequency of exceedance for these example
criteria is no more than once in 3 years, based on recovery rates from literature reviews and
consideration of the life history of insects able to recolonize a site via drift or aerial dispersal (see
Section 3.4). Recovery is expected to occur in 3 years if the following conditions are met:
(1) the SC regime returns to a yearly average below the CCC, (2) there are nearby streams with
low SC supporting a diverse community, and (3) there is an upstream source of colonizers or the
flight or recolonizing distance is within the dispersal range of genetically diverse, reproducing
adult colonizers. If any of these conditions are not met, the time necessary for ecosystem
recovery (and thus, the allowable frequency of exceedance) would likely be longer than 3 years.
4.6. PROTECTION OF FEDERALLY-LISTED SPECIES
Although the derivation of the example criteria was limited to the macroinvertebrate taxa
represented in the data sets, the available evidence indicates that other taxa in the streams would
likely be protected as well (see Section 2.6 and Appendix G). Hence, no adjustment was made
for unanalyzed taxa. However, on a site-specific basis, the example criterion could be adjusted
or recalculated to protect important species, highly valued aquatic communities, or specially
protected waters.
4-27

-------
5. CASE STUDY II: EXAMPLE USING THE EXTIRPATION CONCENTRATION
DISTRIBUTION (XCD) METHOD IN A MODERATE BACKGROUND SPECIFIC
CONDUCTIVITY ECOREGION
This section presents a case study for the Western Allegheny Plateau (Ecoregion 70) to
illustrate how the analytical methods described in Section 3 can be used to derive example SC
criteria using the XCD method in an area with slightly higher background SC than Ecoregion 69
(see Section 4). Ecoregion 70 results, including estimates of the CCC and CMEC, duration,
frequency, and discussion of applicability are included as examples to demonstrate the method.
The derivations of the CCC and CMEC analyses and results are based on data from Ecoregion 70
in West Virginia, and SC data from the Ecoregion 70 outside of WV was used to assess
applicability of the criteria throughout the ecoregion.
5.1. DATA SET CHARACTERISTICS
The Western Allegheny Plateau (Ecoregion 70) extends from the corner of southwestern
Pennsylvania and southeastern Ohio into Kentucky and West Virginia. The hilly and wooded
terrain of the Western Allegheny Plateau is mostly unglaciated and well dissected, with local
relief of 61 to 229 m, and peak elevations of around 610 m. Many of the rivers in this ecoregion
are entrenched, as a result of the rugged, hilly terrain, particularly within the Permain Hills (70a)
and Monongahela Transition Zones (70b). The ecoregion is predominantly forested, but also
consists of a mosaic of urbanized areas, pastures, farms, and coal mines (Woods et al., 1999).
Extensive mixed mesophytic forests and mixed oak forests originally grew in the Western
Allegheny Plateau and, today, most of its rounded hills remain in forest; dairy, livestock, and
general farms, with residential developments concentrated in the valleys. The Western
Allegheny Plateau is composed of horizontally bedded sandstone, shale, siltstone, limestone, and
coal. The horizontally-bedded sedimentary rock underlying the region has been mined for
bituminous coal (Woods et al., 1996).
The data used in this case study are from a large field data set, the WVDEP's in-house
WABbase. Chemical and biological samples are from 1996-2011 and 1997-2010, respectively.
The WABbase contains data from Level III Ecoregions 66, 67, 69, and 70 in West Virginia
(U.S. EPA, 2000a; Omernik, 1987; Woods et al., 1996). The WABbase data set provides
5-1

-------
consistent sampling and analytical methods, high quality, broad spatial coverage of a large
number of perennial streams (2,011 distinct locations) in Ecoregion 70.
The WABbase contains data from a mixed sampling design that collects measurements
from long-term monitoring stations, targeted sites within watersheds on a rotating basin
schedule, randomly selected sample sites (Smithson, 2007), and sites chosen to further define
impaired stream segments in support of TMDL development (WVDEP, 2008a). Most sites are
sampled once during an annual sampling period, but most TMDL sites are sampled monthly for
water quality. The data set contains water quality, habitat, watershed characteristics,
macroinvertebrate data (both raw data and calculated metrics), and geographic location
(WVDEP, 2008a). A wide range of SC levels were sampled, which is useful for modeling the
response of organisms to different ionic concentrations. Level 1 reference status (WVDEP,
2008b) which selects reference sites that "are thought to represent the characteristics of stream
reaches that are minimally affected by human activities and are used to define attainable
chemical, biological and habitat conditions for a region" (WVDEP, 2013). Sites are initially
selected by a map coordinator based on GIS land use data that indicate minimal human
disturbance. Streamside, the appropriateness of the selected site is confirmed based on the level
of anthropogenic disturbance, lack of point discharges, habitat quality, pH, dissolved oxygen,
and SC (>500 (j,S/cm) is used to flag a site for further investigation before inclusion as a
reference site (WVDEP, 2013).
Macroinvertebrate records in the data set are based on collections from a total of 1 m2
area from a 100 m reach at each site. Using a 0.5 m wide rectangular kicknet (595 [j,m mesh),
four 0.25 m2 riffle areas were sampled. In streams narrower than 1 m, nine areas were sampled
with a 0.33 m wide D-frame dipnet of the same mesh size. Composited samples were preserved
in 95% denatured ethanol. A random subsample of 200 individuals (±20%) was identified in the
laboratory. All contracted analyses for chemistry and macroinvertebrate identification followed
WVDEP's internal quality control and quality-assurance protocols (WVDEP 2008b, 2006).
Quality assurance of the data set was judged by EPA to be excellent, based on the database itself
and supporting documentation.
Several data filters, described in Section 3.1 (see Figure 3.2), were applied prior to
finalization of the data set and analyses. A total of 12,909 records from Ecoregion 70 are
included in the data set; of those, SC measurements were included in 11,600 of these samples.
5-2

-------
Many of these are measurements of water quality without biological sampling. Of the 11,600,
there are 2,126 paired samples with SC measurements and biological samples identified to
genus. Of these, a total of 51 samples were removed from the data set due to low pH <6
(48 samples) and high proportion of chloride ions, ([HCO3 ] + [ SO42 ]) < [CI-] (3 samples).
Additional criteria were used to identify macroinvertebrates for inclusion in the extirpation
concentration distribution: occurrence at reference sites and occurrence in 25 or more samples.
Of the 217 macroinvertebrate genera identified in this ecoregion of the WABbase, 179 genera
occurred at least once at one of the 29 identified reference sites where invertebrate samples were
collected and identified to genus. A total of 139 genera occurred at 25 or more sampling
locations. The final example Criterion-data set has 2,075 samples belonging to 1,695 stations (as
depicted in Figure 5-1). Multiple samples were obtained from 19% of stations. Summary
statistics for the data set used to derive the criterion is shown in Table 5-1. The statistical
package R, Version 2.12.1 (December 2010), was used for all statistical analyses
(R Development Core Team, 2011).
5-3

-------
Pennsylvania
Virginia
'# West Virginia
Kentucky
Legend
Sampling stations
Ecoregion 70
Kilometers
Figure 5-1. Ecoregion 70 extends from central Pennsylvania to northern
Tennessee. Sampling sites (stations) (N= 1,695) in the example Criterion data set
that were used to derive the example criterion continuous concentration (CCC)
are indicated as points.
5-4

-------
Table 5-1. Summary statistics of the measured water-quality parameters
used to derive the example specific conductivity criteria in Ecoregion 70.
The example Criterion data set has 2,075 samples belonging to 1,695 stations.
Parameter
Units
Min
25th
50th
75th
Max
Mean3
N
SC
(iS/cm
40
169
259
563
11,646
322.8
2,075
Hardness
mg/L
14.21
67.38
106.44
234.4
2,271.3
130.7
1,050
Total alkalinity
mg/L
1
47.35
84.15
135.75
810
77.3
1,378
SO,2
mg/L
1
19.4
42
253
6,560
67.7
1,405
Chloride
mg/L
0.5
4
7.4
19
1,153
9.4
1,074
SO,2 + HCO,
mg/L
5.7
87.3
168.2
414.4
6,664.9
192.8
1,375
Ca, total
mg/L
1
19.1
30.8
64
621
36.6
1,052
Mg, total
mg/L
1.1
4.64
7.1
15.9
175
9.2
1,053
Na, total
mg/L
1.4
6.4
18.3
52
2,340
22.2
197
K, total
mg/L
0.6
1.2
2.3
4
25.3
2.3
194
TSS
mg/L
1
3
4
7
506
4.5
1,682
Fe, total
mg/L
0.02
0.16
0.3
0.54
137
0.31
1,673
Fe, dissolved
mg/L
0.02
0.02
0.05
0.07
114
0.048
1,285
Al, total
mg/L
0.01
0.09
0.13
0.26
12
0.15
1,392
Al, dissolved
mg/L
0.02
0.02
0.04
0.05
0.4
0.038
1,304
Mn, total
mg/L
0.003
0.02
0.047
0.118
15.9
0.053
1,269
Se, total
mg/L
0.001
0.001
0.001
0.002
0.033
0.002
524
DO
mg/L
1.02
7.89
9.04
10.33
18.35
9.2
2,038
Total phosphorus
mg/L
0.004
0.02
0.02
0.03
2.36
0.023
971
NO2 + 3
mg/L
0.01
0.071
0.1
0.3
30
0.133
971
Fecal
Counts/100 mL
0.5
64
200
581.5
180,000
189
1,955
pH
SU
6.07
7.33
7.64
7.96
10.07
7.6
2,075
Catchment area
km2
0.17
2.88
9.1
38.2
3,912.2
12.2
958
Temperature
°C
0.08
15.9
19.5
22.3
31.9
18.8
2,074
RBPSc
RBP score
49
110
123
136
181
122.9
2,055
RBP_7Sc
Seven most relevant
parameters
31
72
83
94
129
82.8
2,059
Embeddedness
RBP score
0
10
12
14
19
11.4
2,065
Percentage fines
(sand + silt)
Percentage
0
10
20
25
100
20.19
2,033
"All means are geometric means except pH, DO, Temperature, and Habitat variables.
RBP = rapid bioassessment protocol (Barbour et al., 1999; RBP lOSc has 10 parameters while RBP 7 does not
include three flow-related parameters).
5-5

-------
SC ranged from 40-11,646 [j,S/cm, which allowed the response of organisms to be
modeled for a wide range of SC levels. This maximum is SC is three times higher than the
background SC estimated for the data set analyzed in Case Study I (15-3,794 (j,S/cm). Scatter
plots of parameters and SC are depicted in Appendix B.l.
5.1.1. Background Specific Conductivity
Background SC was estimated at the 25th centile from the probability-based samples from
the example Criterion-data set because its sampling design more closely matched the ecoregional
EPA-survey data set. Using this probability-based subsample of the WABbase data set, the
estimated background for Ecoregion 70 was 147 [j.S/cm (681 samples from 617 sites; see
Figure 5-2). Background was also estimated to be 201 [j,S/cm based on field data from
30 reference sites from the WABbase data set (75fe centile; see Figure 5-3). By comparison, the
25th centile for all samples used to derive the example HCos was estimated (166 (j,S/cm) (see
Figure 5-4). The higher estimated background SC based on state-selected reference sites
(n = 30)3 reflects the importance of habitat in site selection and the smaller data set. Seasonal
patterns of SC are evident in the probability-based samples and example Criterion-data set (see
Figures 5-2 and 5-4). The apparent Background SC is <200 [j.S/cm December through June and
>200 [j.S/cm July through October (no samples were available for November; see Figure 5-4).
The effect of seasonal variability of SC on the subsequent analyses was further evaluated and
presented in Appendix B. The large size of the data set and the wide range in SC levels in the
example Criterion-data set allowed for genus XC95 to be calculated.
329 of these sites have biological sampling available as described in Section 5.1.
5-6

-------
HfUUU
5000
f 2000
|
£ 1000
500 -
I 200
100
50
E
Apr
May
Jun
Aug
Sep
Oct
Month
Figure 5-2. Box plot showing seasonal variation of specific conductivity
(jiS/cm) from probability sites from Watershed Assessment Branch database
(WABbase) 1997-2010. This represents a total of 617 sites with 681 samples
from 1997-2010 from Ecoregion 70 with pH >6. Note the difference in scale
along the j'-axis between Figure 5-2 (probability sites) compared to Figure 5-3
(reference sites). See Table 5-2 for sample sizes.
5-7

-------
500
200
Ijn Fib Apr rid
Jus
i	1	r
Jul Aug Sep Dec
Month
Figure 5-3. Box plot showing seasonal variation of specific conductivity
(jiS/cm) in the reference streams of Ecoregions 70 from 1997 to 2010. A total
of 55 samples from 30 reference stations were used for this analysis. Please note
the smaller scale on the_y-axis compared to Figures 5-2 and 5-4. See Table 5-2
for sample sizes.
5-8

-------
t 2000'
i
5 500
i I i i I i i I I i r
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Dec
Month
Figure 5-4. Box plot showing seasonal variation of specific conductivity
(jiS/cm) from all Ecoregion 70 sites from Watershed Assessment Branch
database (WABbase) 1997-2010 used to develop the example criteria.
The example Criterion-data set has 2,075 samples from 1,695 sites. Note the
difference in scale along the j'-axis between Figure 5-2 (all sites, reference and
nonreference) and Figure 5-3 (reference sites). See Table 5-2 for sample sizes.
5.1.2. Ionic Composition
The ionic composition of the samples in the data set for Ecoregion 70 waters was
assessed to ensure that the example criteria were derived for waters dominated by sulfate and
bicarbonate anions (see Figure 5-5). Of the 2,082 samples after low pH samples were removed,
50.3% of samples (1,048 in total) included measures of calcium, magnesium, sulfate,
bicarbonate, and chloride. All but three sites (>99.7%) were dominated by bicarbonate and
sulfate anions, ([HCO? ] + [SO42 ]) > [CI-]. The three chloride-dominated sites were excluded
from the derivation analysis but are shown in Figure 3. Sodium and potassium were less
5-9

-------
frequently measured, but did not exceed calcium and magnesium where measured. Sites with no
ion measurements were retained in the data set because the data had shown that >99.7% of
samples were dominated by bicarbonate and sulfate anions; thus, it is expected that less than 1%
of samples in the Ecoregion 70 Criterion-data set are chloride-dominated. However, the analysis
may also be defensible for ionic mixtures dominated by sodium, sulfate and bicarbonate ions,
e.g., produced water from deep coal mines (Thomas, 2002; Mayhugh and Ziemkiewicz, 2005).
This is because the toxicity of these mixtures are more similar to that of calcium, magnesium,
sulfate and bicarbonate ions than the toxicity of NaCl (Mount et al., 2016; Kunz et al., 2013;
Soucek and Dickinson, 2015).
5.1.3. Seasonal Specific Conductivity Regime
Chemical, physical, and/or biological samples were collected during the sampling years
1997-2010 (January-December). Most sites were sampled once during an annual sampling
period, but some (e.g., sites being studied to improve stream condition within the TMDL
Program) were sampled monthly for water quality parameters (see Table 5-2).
5-10

-------
1unn -
O C'
_i
o
O)
E
C®
I
o
¦Oo o
So o
Co c
100
1000
10
so;- + Hco;tmg/L;
Figure 5-5. Scatter plot of relationship between [CI ] and
([HCO3 ] + [SO42 ]) concentrations in streams of Ecoregion 70 data set. Most
(99.7%) samples (n = 1,045) are below the diagonal line representing the
separation between (HCO3 + SO42 )-dominated and CI -dominated mixtures.
Sites above the 1:1 line were excluded from the example criterion derivation data
set. The Ecoregion 70 data set includes all samples with (HCCV + SO42 ), and
Cl~ measurements. Samples depicted here include all sites regardless of pH.
5-11

-------
Table 5-2. Number of samples with reported genera and specific
conductivity meeting data-inclusion acceptance criteria for the Ecoregion 70
analysis
Number of
Month

samples3
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Total
Full data
set
5
33
11
362
381
290
333
439
193
24
0
4
2,075
Probability
sites
0
0
0
151
262
157
70
17
18
1
0
0
676
Reference
sites
2
3
0
6
15
8
7
3
1
0
0
3
48
Percentage
of total
0.2
1.6
0.5
17.4
18.4
14
16
21.1
9.3
1.2
0
0.2
100
aNumber of samples is presented for each month.
Samples collected from the WABbase-identified reference sites indicate that SC levels
are generally low and similar throughout the year, although slightly higher in summer/fall (see
Figures 5-2 and 5-3). These data show that SC concentrations in flowing waters in the study area
can vary somewhat by season, likely depending on stream discharge, rainfall, snowmelt, and
other hydrological factors. As described in Section 3.1.4 (and in greater detail in U.S. EPA,
201 la), the effects of seasonal differences in SC levels and aquatic insect life history were
evaluated by comparing HCos values partitioned for season. After consideration of the similarity
between the spring HCos and the HCos based on the full data set at the low end of the genus
XCD, the example ecoregional criteria were derived using all available data, regardless of the
time of year they were collected (see Section 3.1.4 and Appendix B.2 in this assessment and
U.S. EPA, 2011a).
5.2. RESULTS
5.2.1. Extirpation Concentration (XC95) and Hazardous Concentration (HC05) Values
(Example Criterion Continuous Concentration)
The Ecoregion 70 example Criterion-data set (see Table 5-1) was used to develop XC95
values from weighted CDFs. The histogram used to develop weights is depicted in Figure 5-6.
The XC95 values that were used in the XCDs are listed in the order of least to most salt-tolerant
5-12

-------
in Appendix B.3. The GAM plots used to designate ~ and > values for those XC95 values are
depicted in Appendix B.4, and the weighted CDFs used to derive the XC95 values used to assign
the XC95 values are shown in Appendix B.5. The example HC05 was calculated at 338 [j,S/cm
(see Figures 5-7 and 5-8); the two-tailed 95% confidence bounds were 272-365 [j,S/cm. Those
bounds, derived by bootstrap resampling, indicate that different data sets could yield HC05 values
within that interval. Rounding to two significant figures, the example CCC is 340 [j.S/cm.
Q
1*0
ou£
-ITT-r^T-rrLn-i
1 1 11 ii 1	1—1—1 1 11111—
100'	1000
"I	1—I I I I I 11—
10000
Specific conductivity (pSIcm)
Figure 5-6. Histograms of the frequencies of observed specific conductivity
values in samples from Ecoregion 70 sampled between 1997 and 2010.
Bins are each 0.017 (1/60) of the range of loglO specific conductivity units wide.
5-13

-------
SB
O
o
*3"
O
CM
o
o
o
338 (J8'cm
500 1000
Specific conductivity (pS/cm)
10000
Figure 5-7. Example genus extirpation concentration distribution (XCD) for
Ecoregion 70. Each point is an extirpation concentration (XC95) value for a genus
(n = 139 genera). The hazardous concentration (HC05) is 338 [j,S/cm
(95% confidence interval 272-365 (j,S/cm) and is the specific conductivity at the
intersection of the genus XCD with the horizontal line at the 5th centile. XC95
values with an approximate or greater than designation are shown as triangles.
5-14

-------
100	200	500	1000
Specific conductivity (liStem)
Figure 5-8. The lower end of the example genus extirpation concentration
distribution for Ecoregion 70. The dotted horizontal line is the 5th centile. The
vertical arrow indicates the hazardous concentration (HCos) of 338 [j,S/cm (95%
confidence interval 272-365 (j,S/cm). Only the 50 most salt-intolerant genera are
shown to better discriminate the points on the left side of the distribution. The six
most salt-intolerant genera (i.e., extirpation concentration [XC95] < 338 (j,S/cm)
are Drunella, Utaperla, Cinygmula, Alloperla, Ephemerella and Heptagenia.
XC95 values with an approximate or greater than designations are shown as
triangles.
5.2.2. Example Criterion Maximum Exposure Concentration
At sites meeting the CCC of 340 [j,S/cm, 90% of the SC observations are estimated to
occur below the CMEC (see Section 3.2). The CMEC was derived using the Ecoregion 70 data
set. Out of the 12,909 samples collected between 1996-2011, 8,302 samples had a July-to-June
rotating year representing 819 rotation years and 805 unique stations, with at least 1 sample from
July to October and one from March to June, and at least 6 samples within a rotation year with
5-15

-------
SC measurements (see Table 5-3). Note that inclusion of samples is not contingent on biological
data. Reference and nonreference sites were included to ensure a range of SC (see Table 5-1).
Table 5-3. Summary data related to the calculation of the example criterion
maximum exposure concentration (CMEC) for Ecoregion 70
Number of samples June to July prior to biological sampling
8,302
Number of rotation years (# unique stations)
819 (805)
Number of WVDEP reference sites
12
ccc
340 (j,S/cm
CMEC
680 (j,S/cm
Of the 819 rotation years (805 unique stations) with multiple SC measurements, the
variability of within station SC was found to differ among streams (see Figure 5-9); however, the
LOWESS line indicated that the average variability (residual standard deviation for a station) is
not very different across the entire gradient in Ecoregion 70. The stations with annual mean SC
between the 25th and 75fe centile (120 and 520 (j,S/cm) were used to estimate the variance
components of annual mean SC (at 340 (j,S/cm). The selected data sets with mean SC values
between 120 and 520 [j,S/cm in respective data sets have a sample size of 518 rotation years
(513 stations) and 5,272 observations. The grand mean and standard deviation of this data set
were determined and the CMEC was calculated. The CMEC calculation is shown below:
CMEC for Ecoregion 70: iolog10(340)+ 128*0.237 = 6g4 ^s/cm
The example field-based calculated CMEC rounded to two significant figures yields a
CMEC of 680 [j,S/cm for Ecoregion 70. At this level, where the annual average SC <340 [j,S/cm,
90% of the observations are expected to be less than the CMEC.
5-16

-------
o
10U	1U0U	10000
Station mean specie conductivity
Figure 5-9. Illustration of within site variability (residual standard deviation
for each station) along the specific conductivity gradient (station mean) in
Ecoregion 70. The x-axis is log annual mean specific conductivity. Each dot
represents a station. The fitted line is the locally weighted scatterplot smoothing
(LOWESS, span = 0.75, linear polynomial model), while the two vertical dashed
lines represent logarithm mean specific conductivity of 120 and 520 [j,S/cm
respectively. Within those bounds, the standard deviation is fairly constant.
5.3. GEOGRAPHIC APPLICABILITY
The geographical applicability of the example criteria throughout Ecoregion 70 was
assessed using the background-matching approach (see Section 3.7.1). The background SC of
the new area (i.e., Ecoregion 70 beyond West Virginia) was estimated at the 25th centile (see
Section 3.7.1.2; and Cormier and Suter, 2013a) and compared with the background estimates for
Ecoregion 70 within West Virginia.
5-17

-------
Because the example SC criteria presented here have been developed for a dissolved
mixture dominated by sulfate and bicarbonate anions ([HCO3 ] + [SO42 ]) > [CP] in mg/L), all
chloride-dominated samples ([HCO3 ] + [SO42]) < [CP] in mg/L) were removed from the data
set before estimating background SC. Thereby, the background for the new area is estimated for
the same ionic mixture as the example criteria.
5.3.1. Data Sources
Two data sets were used for this example applicability assessment: the original data set
used to derive the HC05 described in Section 5.1 and an EPA-survey data set.
An EPA-survey data set was used to evaluate and characterize ion concentrations and
water chemistry in the ecoregion. The primary sources of the combined data are from EPA
survey programs: the NRSA 2008-2009 data set (U.S. EPA, 2013b), WSA 2004 data set
(U.S. EPA, 2006), EMAP 1993-1998 data sets and R-EMAP 1999 data set (U.S. EPA, 2013c),
and NAPAP data set collected in 1986 (NADP, 2013) (see Table 5-4). Data sets are based on
single random samples from June through September. Most report SC, alkalinity, hardness,
sulfate, chloride, bicarbonate, pH and other water quality parameters. Ecoregions and sampling
sites are shown in Figure 5-10. All samples were collected from first-through fourth-order
streams as part of a probability-based design intended to estimate proportions of parameters for
various stream classes. The probability-design weights were not used in this characterization.
Analysis of water chemistry samples followed EPA procedural and quality assurance/quality
control protocols from EMAP (U.S. EPA, 2001, 1998b, 1994, 1987), Wadeable Streams
Assessment (U.S. EPA, 2004a, b), NRSA (U.S. EPA, 2009), and NAPAP (Drouse et al., 1986;
U.S. EPA, 1987). These data sets were also selected so that methods would be comparable
across the data set and because these studies used probability-based designs (i.e., randomly
assigned sampling locations).
5-18

-------
Table 5-4. Description of survey data sets combined to form the EPA-survey
data set used to assess applicability of example ecoregional criteria
throughout Ecoregion 70
Data set
Sampling period
Total TV
KY
OH
PA
MAHA EMAP
1993-1995
14
0
0
14
MAIA EMAP
1997-1998
10
0
0
10
WSA
2004
16
5
6
5
NRSA
2008-2009
14
4
6
4
NAPAP
1986
5
0
0
5
Region 4 Wadeable Streams R-EMAP
1999-2002
2
2
0
0
Total

61
11
12
38
/
•!«
• *•
"m
: ennsyivama
Kentucky •	•
Original
area
Legend
Virginia
• EPA combined data (Eco70)
i Ecoregion70
0 1530 60 90 120
¦ Kilometers
Figure 5-10. Ecoregion 70 extends from southwestern Pennsylvania and
southeastern Ohio into Kentucky. Sampling sites in the EPA-survey data set
that were used to estimate background in the "new" area for Ecoregion 70 are
indicated as points. There are 61 samples from 61 stations.
5-19

-------
5.3.2. Geographic Applicability Results
A summary of water quality for the ecoregion, including major ionic constituents, for the
EPA-survey data set is listed in Table 5-5. Sites with HCO3 + SO42 concentrations on a mass
basis greater than or equal to CP were used to estimate background SC. This mixture is common
in the ecoregion, and only one site was dominated by chloride anions in the EPA-survey data set
and none in the example Criterion-data set. Therefore, this one site was excluded so the natural
background was estimated from the altered EPA-survey data set.
Background SC for bicarbonate and sulfate dominated waters estimated as the 25th centile
of the EPA-survey data set for the area in Ecoregion 70 outside the area used to develop the
example criteria was 197 [j.S/cm (95% CI 145-272 (j.S/cm) (see Table 5-6). The 25th centile
from the probability sample from the example Criterion-data set was 147 [j,S/cm
(95% CI 136-159) (see Table 5-6). The confidence bounds for background estimated from the
example Criterion-data set overlap with the confidence bounds for background estimated for the
rest of the ecoregion. Therefore, the background SC regime throughout Ecoregion 70 appears to
be similar, and the example criteria (CCC = 340 [j,S/cm, CMEC = 680 (j,S/cm) are considered
geographically applicable throughout the ecoregion. Other estimates of background from the
reference sites in the example Criterion-data set (201 [j,S/cm; 95% CI 164-210 (j,S/cm) and the
example Criterion data set (169 [j,S/cm; 95% CI 161-171 (j.S/cm) also overlap with the CI of the
background for the rest of the ecoregion. As a validation of background specifically for the
portion of Ecoregion 70 in Ohio, a weight of evidence was performed (see Appendix C).
5-20

-------
Table 5-5. Summary of water quality parameters for Ecoregion 70
EPA-survey data set
Ion or specific
conductivity
Min
Centile
Max
aN
10th
25th
50th
75th
HCO3 (mg/L)
0.0
1.8
18.1
52.2
121.0
241.8
42
S042 (mg/L)
11.5
17.2
24.1
52.6
144.7
955.8
58
cr (mg/L)
1.0
3.8
6.0
9.6
26.3
204.5
58
Ca2+ (mg/L)
4.9
10.5
19.0
47.8
69.3
240.8
58
Mg2+ (mg/L)
1.8
3.5
6.2
12.5
22.7
87.7
58
Na+ (mg/L)
1.0
3.1
4.3
9.1
22.8
161.2
58
K+ (mg/L)
1.0
1.6
1.9
2.4
3.0
9.6
58
pH (SU)
4.0
6.3
7.3
7.7
8.1
8.6
60
b(HC03 + so42 )/cr
1.7
3.7
8.0
10.6
22.8
103.8
42
SC (nS/cm)
66.7
108
197
398
631
1,860
60
aRelevant N indicates the number of samples from the large data set relevant to each water quality parameter.
bValue within category calculated from individual sample ion concentrations. HCO:, + SO i2 /CI in mg/L greater
than 1 indicates the mixture is dominated by HCO;, + SO r . One site dominated by CI" was removed from the
data set.
Table 5-6. Background specific conductivity estimates for Ecoregion 70
Data set
Centile used to
estimate
background
Estimated
background
jiS/cm
Confidence
interval
jiS/cm
Relevant N
(stations/
samples)
EPA-survey data set from geographic
area in Ecoregion 70 not represented
in the example criterion derivation
data set
25th
197
135-240
60/61
WABbase data set, probability sample
subset
25th
147
136-159
617/681
WABbase data set, reference sample
subset
75th
201
164-210
30/55
Ecoregion 70 example criterion
derivation data set, full data set
25th
169
161-171
1,695/2,075
5-21

-------
5.4.	SUMMARY OF EXAMPLE CRITERIA FOR ECOREGION 70
The case example for Ecoregion 70 includes an annual geometric mean (i.e., CCC) and a
1-day mean (i.e., CMEC), not to be exceeded more than once in 3 years on average. Both of
these distinct expressions of the example SC criteria would need to be met in order to adequately
protect aquatic life. These values indicate that freshwater animals would be protected if the
annual geometric mean SC concentration does not exceed 340 [j,S/cm and the 1-day mean does
not exceed 680 [j,S/cm, more than once every 3 years on average. These example criteria would
apply to all flowing freshwaters (ephemeral, intermittent, and perennial streams) in Ecoregion 70
inclusive of portions of Kentucky, West Virginia, Pennsylvania, and Ohio. On a site-by-site
basis, these example ecoregional criteria apply if the ionic mixture is dominated by anions of
bicarbonate and sulfate. For streams crossing into Ecoregion 70, professional judgment may be
needed to assess the potential effect of different ionic composition or concentration. Professional
judgment is recommended when applying to sites with a catchment area greater than 1,000 km2
(386 mi2) owing to lesser representation in the example data set by this class of stream. On a
site-by-site basis, alternative SC criteria may be more appropriate if the natural background of a
site is shown to be lower or higher than its regional background specific conductivity.
5.5.	EXAMPLE CRITERION CHARACTERIZATION
5.5.1. Factors Potentially Affecting the Extirpation Concentration Distribution (XCD)
Model
An assessment of potential confounders and an analysis of the influence of habitat quality
and sampling date for Ecoregion 70 can be found in Appendix B.2.
5.5.1.1. Sensitivity Analyses
As the minimum number of occurrences of a genus for inclusion in the data set increases,
fewer genera are included in the XCD. The HCos increases greatly when a taxon in the lower
5th centile is removed because it does not meet the minimum number of samples and then more
slowly alternates between increasing and decreasing as genera either above or below the
5th centile are removed because they do not meet the minimum number of samples (see
Figure 5-11). The pattern repeats until all genera above and below the lower 5th centile have the
same XC95 value (not shown). To maximize the number of genera included in the XCD, a
minimum of 25 occurrences was utilized.
5-22

-------
The number of samples in the data set affected the number of genera included in the XCD
and the resulting example HCos. The effects of data set size on the HCos estimates and on their
confidence bounds were estimated using a bootstrapping technique. The mean of all
bootstrapped HCos values, the numbers of genera used for the HCos calculation, and their 95% CI
were plotted to show the effect of sample sizes. As shown in Figure 5-12, the HCos for this data
set stabilizes, reaching an asymptote at approximately 500-800 sites sampled and 90-100 genera
evaluated. Therefore, the original data set was considered adequate for estimating the example
ccc.
o
CO
=L
&
O
=3
TJ
a
o
o
Cl
1)
Cl
00
iO
d3
X
340 -
320 -
300
280 -
260 -
10 20 30 40 50 60
Minimum observations for inclusion of genus
Figure 5-11. Relationship of the number of occurrences of a genus on the
hazardous concentration (HCos) based on Ecoregion 70 example
Criterion-data set. Estimates of HCos values (blue diamonds, left axis) and the
number of taxa in the extirpation concentration distribution (XCD) (red squares,
rights-axis) based on minimum number of samples (5-60, x-axis). As the
minimum number of occurrences of a genus increases, fewer are included in the
XCD.
5-23

-------
900 -
800
i 700
uo
co 600
 500
LD
O
U
X
400 "
300 -
I
I

500
1000	1500
Sample Size
2000
- 141
122 O
U
X
c
- 103
- 84
- 55
- 46
- 28
Figure 5-12. Adequacy of the size of the data set used to model the
hazardous concentration (HCos) based on the Ecoregion 70 example
Criterion-data set. As size of the data set increases, the number of genera
included in the genus extirpation concentration distribution (XCD) increases
(triangles). The HCos stabilizes, reaching an asymptote at approximately
500-800 sites sampled (circles) and 90-100 evaluated genera.
ra
s_
CD
C
QJ
dj
_Q
E
5.5.2. Validation of the Model
As recommended by the SAB (U.S. EPA, 201 lc), the XCD model was validated and
uncertainty around the HCos values was estimated using bootstrapping. The median HCos
estimated from bootstrapping was 323 jj,S/cm (95% CI 272-365 pS/'cm) which is similar to the
HCos of 338 (aS/cm measured using a two-point interpolation from the original XCD. The
similarity between the two HCos values indicates a similar model would be generated using an
independent data set (see Figure 5-13).
5-24

-------
o 	1	1	'	1	1	'	'—<~~T~
200	500	1000
Specific conductivity (ijS/cm)
Figure 5-13. 95% confidence intervals (hatched oblique lines) for the lower
portion of the Ecoregion 70 genus extirpation concentration distribution
(XCD). Each tiny gray dot represents an extirpation concentration (XC95) value
from one of 1,000 XCD bootstrapping iterations (note that the genera and their
order varies with each XCD-iteration). Each of the 36 blue filled dots represents
the calculated XC95 of the XCD for the example criterion continuous
concentration (CCC). Hazardous concentration (HC05) based on the bootstrap
medians is 323 (iS/cm.
5.5.3. Duration and Frequency
Numeric criteria include magnitude (i.e., how much), duration (i.e., how long), and
frequency (i.e., how often) components. Appropriate duration and frequency components of
criteria are determined based on consideration of available data and understanding the
exposure-response relationship in the context of protecting the aquatic life of a water body. The
significant consideration used in setting the duration component of aquatic life criteria is how
long the exposure concentration can be above the criteria without affecting the endpoint on
which the criteria are based (U.S. EPA, 1985, 1991). Based on the temporal resolution of the
available field data set and an analysis of within-site variability of SC levels, EPA developed two
5-25

-------
different expressions for the example SC criteria in order to provide adequate protection for
aquatic life.
In this case, the majority (>81%) of sites used to derive the example CCC for
Ecoregion 70 were sampled once during an annual sampling period and thus represent the
average stream chemistry (SC) and macroinvertebrate assemblage information over the course of
1 year. As a result, the appropriate duration for the CCC is 1 year. The duration for the CMEC,
a level of protection from acutely toxic exposures, is 1 day based on a review of the literature on
the onset of macroinvertebrate drift in response to elevated SC (see Section 3.3). At sites
meeting the CCC, 90% of the SC observations are estimated to occur below the CMEC.
EPA anticipates that an appropriate allowable frequency of exceedance for these example
criteria is no more than once in 3 years, based on recovery rates from literature reviews and
consideration of the life history of insects able to recolonize a site via drift or aerial dispersal (see
Section 3.4). Recovery is expected to occur in 3 years if the following conditions are met:
(1) the SC regime returns to a yearly average below the CCC, (2) there are nearby streams with
low SC supporting a diverse community, and (3) there is an upstream source of colonizers or the
flight or recolonizing distance is within the dispersal range of genetically diverse, reproducing
adult colonizers. If any of these conditions are not met, the time necessary for ecosystem
recovery (and thus, the allowable frequency of exceedance) would likely be longer than 3 years.
5.6. PROTECTION OF FEDERALLY-LISTED SPECIES
Although the derivation of the example criteria was limited to the macroinvertebrate taxa
represented in the data sets, the available evidence indicates that other taxa in the streams would
likely be protected as well (see Section 2.6 and Appendix G). Hence, no adjustment was made
for unanalyzed taxa. However, on a site-specific basis, the example criterion could be adjusted
or recalculated to protect important species, highly valued aquatic communities, or specially
protected waters.
5-26

-------
6. CASE STUDY III: EXAMPLE USING THE BACKGROUND TO CRITERION (B-C)
REGRESSION METHOD
This section presents an example calculation of an ecoregional CCC using the B-C
method (see Section 3.7.2 and Appendix D). In this example, a CCC for the Northwestern Great
Plains, Level III Ecoregion 43, was calculated using SC data from the ecoregion and the B-C
method because there were insufficient paired SC and biological data to use the XCD method in
this ecoregion.
First, the water chemistry data set was screened for ionic composition to ensure samples
were dominated by sulfate and bicarbonate anions, and sampled sites were mapped to determine
whether the geographic distribution of sites was representative of (dispersed throughout) the
ecoregion. Minimally affected background SC of the new ecoregion was estimated at the
25th centile of probability samples (see Section 3.7.1.2; Cormier and Suter, 2013a). Least
disturbed background SC was estimated at the 25th centile of a combined data set of targeted and
probability samples. The CCC was calculated using the least disturbed 25th centile background
SC as the independent variable (x) in the B-C regression model to yield an HCos (>')• Depending
on available data and analytical results (see Figure 3-11), an HCos may take the form of (1) the
value at the mean of the regression line from the B-C mode, (2) the value at the lower 50%
PL of the regression line, or (3) an HCos derived from a data set based on >200 paired SC and
biological samples. In this example case for Ecoregion 43, there were <200 paired SC and
biological samples, so the lower 50% PL was used to develop the example CCC.
6.1. DATA SET CHARACTERISTICS
The Northwest Great Plains is mostly an unglaciated, semiarid, rolling plain with rolling
hills and occasional buttes and badlands (Woods et al., 2002). Elevation ranges from 458 to
1,200 meters (McNab and Avers, 1994). The area covers approximately 347,000 km2 and
extends from southeastern Montana and northeastern Wyoming into western parts of North
Dakota and South Dakota. Ecoregion 43 is bordered by the Northwestern Glaciated Plains to the
north and east, the Middle Rockies and Wyoming Basin to the west, the Eastern High Plains and
Nebraska Sand Hills to the south. An outcropping of the Middle Rockies occurs in the south of
the ecoregion. The shallow soil is underlain with shale, siltstone, and sandstone. Where there is
6-1

-------
sandstone, aquifers can produce groundwater. Otherwise, there are few perennial rivers, and the
rainfall is erratic with approximately 250-510 ml/year. The low precipitation and high
evapotranspiration lead to less groundwater recharge and baseflow contributing to streams;
therefore, many small streams are intermittent or ephemeral. Grazing and ranching is a common
land use with some dryland and irrigated agriculture. Surface coal mining and oil and gas
production also occur. The often alkali-rich soils in the steppes are dominated by sagebrush;
whereas, the buttes are more moist and can support forests.
Only existing data were used for this example assessment (see Table 6-1). An
EPA-survey data set and a U.S. Geological Survey (USGS) data set were combined to
characterize ion concentrations and water chemistry and then used to calculate a provisional
CCC. The USGS data set was also used to calculate the CMEC. The statistical package R,
Version 2.12.1 (December 2010), was used for all statistical analyses (R Development Core
Team, 2011).
Table 6-1. EPA and U.S. Geological Survey (USGS) chemistry data sets
included in this study.
Years indicate the period during which the samples were collected. Western
Environmental Monitoring and Assessment Program (EMAP) survey sites are
included in the count of sites from the National Wadeable Streams Assessment
(NWS A).
Survey
Years
# of sites
# of samples
EPA probability samples
EMAP and Regional EMAP
1993-2003
12
12
NWS A
2000-2004
53
53
NRSA
2008-2009
53
53
USGS mixed sampling
USGS: full data set
1946-2008
281
45,489
USGS: subset >_six samples per rotation year, July-June
1946-2008
148
41,648
This B-C regression model was developed using biological data paired with SC data from
24 data sets with waters having ionic mixtures dominated by calcium, magnesium, sulfate and
bicarbonate ions and where background SC did not exceed 626 [j,S/cm. Therefore, the model is
6-2

-------
most appropriate for waters with similar ionic characteristics. The model has not been
thoroughly tested with waters dominated by other mixtures, i.e., ([SO42 ] + [HCO3 ]) < [cr],
and ([Ca2+] + [Mg2+]) < ([Na+] + [K+]) in mg/L. In particular, the B-C model is not appropriate
for waters dominated by NaCl (Haluszczak et al., 2013, Entrekin et al., 2011; Gregory et al.,
2011; Veil et al., 2004) or road salt (Forman and Alexander, 1998; Kelly et al., 2008;
Environment Canada and Health Canada, 2001; Evans and Frick, 2001; Kaushal et al., 2005).
However, the model and this analysis may be defensible for ionic mixtures dominated by
sodium, sulfate and bicarbonate ions (Brinck et al., 2008; Dahm et al., 2011; Jackson and Reddy,
2007; National Research Council, 2010; Clark et al., 2001; Veil et al., 2004). This is because the
toxicity of these mixtures are more similar to that of calcium, magnesium, sulfate and
bicarbonate ions than the toxicity of NaCl (Mount et al., 2016; Kunz et al., 2013; Soucek and
Dickinson, 2015).
In this example case study, more than half of the sampled sites were dominated by sulfate,
bicarbonate, sodium, and potassium ions, ([SO42 ] + [HCO3 ]) > [Cl~], and
([Ca2+] + [Mg2+]) < ([Na+] + [K+]) in mg/L. No samples were excluded based on cations. All
samples in the EPA-survey data set were used because no samples were dominated by chloride
ions. A USGS data set of 281 sites was used to verify ionic composition. Of 7,461 samples,
7,456 (>99.9%) were dominated by sulfate plus bicarbonate. The five samples not dominated by
sulfate and bicarbonate occurred at sites sampled multiple times that more often than not were
dominated by sulfate and bicarbonate, so these sites were retained. All sites in the combined
EPA-USGS data set had pH data, but none were <6 nor >9.8, so no sites were removed from the
data set.
6.1.1. EPA-Survey Data Set
Data sources, sampling period, and number of samples used to estimate background SC
in the new area (Ecoregion 43) are provided in Table 6-1. The NRSA 2008-2009 data set
(U.S. EPA, 2013b), WSA 2004 data set (U.S. EPA, 2006), EMAP 1993-1998 data sets and
R-EMAP 1999 data sets (U.S. EPA, 2013c), are based on single random (i.e., probability-based
design) samples from June through September.
EPA-survey data sampling sites within the ecoregion are shown in Figure 6-1. Water
quality parameters collected in Ecoregion 43 are included in Table 6-2. Most of the samples
6-3

-------
have reported SC, alkalinity, hardness, sulfate, chloride, bicarbonate, and pH, as well as other
water quality parameters. When necessary, ionic concentrations in milliequivalents (meq/L)
were converted to mg/L [(meq/L) x (ion MW)/(ionic charge)] (Hem, 1985).
MT
ND
SD
"~A
WY
NE
200
400 km v
—i	\
S
Figure 6-1. Sampling sites in the EPA-survey data set that were used to
estimate minimally affected background in Ecoregion 43. For the purpose of
demonstrating the background-to-criterion approach, Level III Ecoregion 43
which encompasses portions of Montana, North and South Dakota, Wyoming,
and Nebraska is shaded gray. A total of 115 sampling sites is depicted. Sampling
locations are color-coded by site-specific conductivity range: green diamonds
<300 [j,S/cm, yellow squares 300-1,000 [j,S/cm, and red triangles >1,000 [j,S/cm.
Geodetic reference system = North American Datum (NAD83).
6-4

-------
Table 6-2. Summary of data for the example case for Northwestern Great
Plains, Level III Ecoregion 43 from EPA-survey samples.
Geometric means were calculated except for pH and ion ratio.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
118
57
483
1,257
1,041
2,406
5,769
pH (SU)
118
6.70
8.06
8.27
8.30
8.47
9.88
Ca2+ (mg/L)
111
3.62
37.6
53.5
62.6
139
511
Mg2+(mg/L)
111
0.65
9.34
33.0
26.8
82.1
240
Na+ (mg/L)
106
1.23
23.7
162.97
99.2
390
1,059
K+ (mg/L)
106
0.44
4.16
7.95
6.73
12.6
80.1
HCO3 (mg/L)
53
45.8
205
277
286
429
987
S042 (mg/L)
106
3.33
52.4
367
214
1,074
2,750
cr (mg/L)
106
0.20
3.80
8.50
8.63
18.9
520
a hco3 + so42 /cr
53
3.26
35.2
73.3
99.4
129
464
aRatio of mg/L HCO3 + SO42 /CI greater than 1 indicates the mixture is dominated by HCO3 + SO42 .
All samples were collected from first-through fourth-order streams as part of a
probability-based design intended to estimate proportions of parameters for various stream
classes. The probability-design sampling weights for stream order were not used in the
characterization. Analysis of water chemistry samples followed procedural and QA/QC
protocols of EPA andEMAP (U.S. EPA, 2001, 1998b, 1994, 1987), Wadeable Streams
Assessment (U.S. EPA 2006, 2004a, b), NRSA (U.S. EPA, 2009), and NAPAP (Drouse et al.,
1986; U.S. EPA, 1987).
6.1.2. U.S. Geological Survey (USGS) Data Set
The USGS survey data set used in this example is composed of 45,489 water quality
samples, from 281 stations within Ecoregion 43 (see Table 6-1; Figure 6-2). Some stations were
sampled only once while others were sampled as many as 5,445 times. The data were collected
between 1946 and 2015 during all seasons. Water quality parameters collected in Ecoregion 43
are included in Table 6-3. Most of the samples have reported SC, alkalinity, hardness, sulfate,
chloride, bicarbonate, and pH, as well as other water quality parameters. Analysis of water
6-5

-------
chemistry samples followed procedural and QA/QC protocols for USGS data sets (Mueller et al.
1997).
MT
ND
~ AJ
~ ~
SD
W
WY
200
400 km
Figure 6-2. Sampling sites in the U.S. Geological Survey (USGS) data set in
Ecoregion 43. Ecoregion 43 is shaded gray. Geometric mean specific
conductivity at sampling locations is color-coded: green diamonds <300 [j,S/cm,
yellow squares 300-1,000 [j,S/cm, and red triangles >1,000 [j,S/cm. Geodetic
reference system = North American Datum (NAD83).
6-6

-------
Table 6-3. Summary of data for the example case for Northwestern Great,
Level III Ecoregion 43 from U.S. Geological Survey (USGS).
Geometric means were calculated for all variables except for pH and ion ratio.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
281
85
564
1,045
986
1,816
7,330
pH (SU)
170
7.30
8.1
8.2
8.2
8.3
9.0
Ca2+ (mg/L)
118
21.87
50.9
69.0
76.4
114.9
464
Mg2+ (mg/L)
118
5.81
28.4
46.6
45.0
75.5
654
Hardness (mg/L)
52
121.1
290
479.7
432.4
608.1
1,040
Na+ (mg/L)
118
1.55
59.4
180.3
132.5
325
1,186
K+ (mg/L)
110
0.74
5.1
9.2
8.1
13.3
25.1
HCO3 (mg/L)
92
2.00
224.4
278.5
204.2
405
765
S042~(mg/L)
120
4.56
189.5
464.2
362.4
808.3
2,283
Cr (mg/L)
117
0.97
6.9
10.7
14.7
27.6
938
8hco3 + so42 /cr
86
2.35
22.9
77.6
82.6
116
363
aratio of mg/L HCO3 + SO42 /CI greater than 1 indicates the mixture is dominated by HCO3 + SO42 .
All samples were collected from first-through fourth-order streams for various research
purposes and thus the targeted sampling design may emphasize areas with increased
anthropogenic disturbance or some geologic formations. In this respect, the data may skew the
background SC estimates. Background SC was not estimated from this data set because
sampling stations were not randomly assigned. However, after weighing the potential bias that
could be introduced with the benefits of having greater coverage across the ecoregion, the
EPA-survey and USGS data sets were combined and used to estimate background SC.
Because the USGS data set contained multiple measurements in an annual rotation from
sampling locations, the data set was used to estimate a CMEC and to explore the variability of
SC patterns within the region. Therefore, a second data set was selected by excluding stations
with fewer than six SC measurements throughout the year. A minimum of one sample during
the first 6 months and one in the last 6 months of the previous year were also required so that at
least one low and high SC sample was included in the data set. The second USGS data set that
included 41,648 samples from 168 stations was used to calculate the variance near the CCC for
the CMEC calculation.
6-7

-------
6.1.3. Modeled Mean Baseflow Background Specific Conductivity (SC)
Predicted mean natural base-flow SC in catchments of Northwestern Great Plains,
Ecoregion 43, was also considered for comparison purposes (Olson and Hawkins, 2012). The
stream length weighted, mean natural SC (each SC was multiplied by the proportion of stream
segment length) at base flow for each ecoregion (see Appendix D). Figure 6-3 shows the
predicted SC at 300m resolution in order to emphasize the general trends across Ecoregion 43.
MT
ND
Specific
conductivity (nS/cm)
600
500
400
,SD
300
200
100
WY

150
300 km
Albers Equal Area Conic USGS version
c
Figure 6-3. Predicted mean natural base-flow specific conductivity in
catchments of Northwestern Great Plains, Ecoregion 43, using the
Olson-Hawkins model. Albers projection used for mapping.
6.1.4. Characterization of Ionic Matrices
6.1.4.1. EPA-Survey Data Set Ionic Characteristics
A summary of water quality ionic constituents including major ionic constituents for
Ecoregion 43 is provided in Table 6-2. Centiles were calculated using each sample observation
6-8

-------
because most measurements were single grab samples. There were no chloride-dominated sites
where ([CP] > [HCO3 ]) + [SO42 ] in the EPA-survey data (A' = 118); therefore, no sites were
excluded from the data sets. Sodium was the dominant cation at more than half the sampled
sites. This mixture was judged as acceptable for use with the B-C model, but with less
confidence than a calcium and magnesium dominated mixture. Ionic characteristics for
Ecoregion 43 are shown for the EPA-survey in Table 6-2 and the USGS data set in Table 6-3.
6.1.4.2. U.S. Geological Survey (USGS) Data Set Ionic Characteristics
Table 6-3 summarizes the water quality parameters including major ionic constituents for
Ecoregion 43 in the USGS data set. Unlike the EPA-survey data set, the USGS data set contains
targeted sites of interest rather than probability samples. Also, in some cases, there are multiple
measurements from the same site, and other sites are autocorrelated with downstream sampling
locations. Therefore, the distribution of sites in this data set is not necessarily representative of
Ecoregion 43 in its entirety; however, the data set can be used to define the overall pattern of SC
for the ecoregion because it contains samples in areas not represented in the EPA-survey data set.
Centiles were calculated using the geometric mean of site measurements (except pH and the
ionic ratio) and were qualitatively compared to the probability-based EPA-survey data set.
Two samples with pH <6 and a few observations (less than 10) with some ion
concentrations recorded as 0 were removed from the USGS data set. Only 5 out of
45,489 samples collected in this data set were dominated on some days by chloride,
[CP] > ([HCO3 ] + [SO42 ]) in mg/L; however, on average, these 5 sites were not dominated by
chloride ions, so those samples were not removed. Sodium was the dominant cation in more
than half of the samples. This mixture was judged as acceptable for use with the B-C model, but
with less confidence than a calcium and magnesium dominated mixture.
6.1.5. Comparison of Background Specific Conductivity (SC) Estimates
In this case example, the stream length weighted average predicted mean base flow SC
from the Olson-Hawkin's model in Ecoregion 43 is 489 [j,S/cm. Raw values for predicted mean
flow for stream segments are shown as box plot in Figure 6-4. The 25th centile of the
EPA-survey data set is 483 [j,S/cm. The USGS SC data set has a slightly narrower overall range
and mid-range of values resulting in a slightly higher quartile SC than the EPA survey data set
6-9

-------
(564 (j,S/cm) (see Figure 6-4 and Table 6-3). When the USGS data set is combined with the
EPA-survey data set, the 25th centile (542 (j,S/cm) is also greater than the predicted mean
baseflow of 489 [j,S/cm (see Table 6-4, Figure 6-4). Therefore, this background SC estimated
from the combined EPA-USGS data set is least disturbed.
The 25th centile of the combined EPA-survey and USGS data set was used to calculate
the HCos following the decision tree described Section 3.7.2 and Figure 3-11. Because the lower
quartile ranges from 85 to 564 |iS/cm, additional analysis is recommended for streams known to
have low SC regimes.
5000 -
2000 -
IT
0
^.1000 -
£
1	500 -
C
o
o
o
£=
£ 200 -
CO
100 -
50 -
Figure 6-4. Box plots of specific conductivity (SC) distributions for
EPA-survey, U.S. Geological Survey (USGS), combined EPA-survey and
USGS data sets and predicted mean base-flow. The USGS data set captures a
slightly narrower midrange of values possibly owing to the targeted sampling and
the mean values rather than the single measurements in the EPA sample. The
25th centile of the combined and USGS data set is greater than the mean predicted
base-flow. The mean baseflow model predicts many outliers for the region
<200 |iS/cm.
Baseflow
Combined
EPA-survey
USGS
6-10

-------
Table 6-4. Summary of data for the example case for Northwestern Great
Plains, Level III Ecoregion 43 from the combined EPA-survey and
U.S. Geological Survey (USGS) data set.
Geometric means for sampled sites were calculated for all variables except for pH
and ion ratio.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
399
57
542
1,074
1,002
2,006
7,330
pH (SU)
288
6.7
8.1
8.2
8.3
8.4
9.9
Ca2+ (mg/L)
229
3.6
46.2
63.0
69.4
122.5
511.2
Mg2+ (mg/L)
229
0.65
20.0
43.7
35.0
77.6
654.4
Hardness (mg/L)
163
11.7
184.5
377.8
325.0
642.5
1,828.1
Na+ (mg/L)
224
1.2
49.1
173.7
115.5
337.7
1,185.6
K+ (mg/L)
216
0.44
4.5
8.7
7.4
12.9
80.1
HCO3 (mg/L)
145
2.0
212.9
277.8
231.0
410.8
986.7
S042 (mg/L)
226
3.33
129.5
452.9
283.0
940.8
2,750.7
CP (mg/L)
223
0.20
5.3
9.9
11.4
20.6
937.8
aHC03 + so42 /cr
139
2.35b
33.6b
76.25b
89b
123.0b
464b
aRatio of mg/L HCO3 + SO42 /CI greater than 1 indicates the mixture is dominated by HCO3 + SO42 .
6.1.6. Calculation of Ecoregion 43 Criterion Continuous Concentration (CCC) from
Background
Because available paired SC and biological data represented <200 sites, the HC05 was
estimated at the lower 50% PL using the B-C model (see Figure 6-5). The 25th centile of SC
from the EPA-survey data set of Ecoregion 43 was used to identify the lower 50% PL using
eqs 3-4 and 3-5. The B-C model development is described in Appendix D. The x-variable is the
background SC in Ecoregion 43 which was loglO transformed. The calculated .y-value is the
predicted mean loglO HC05. In this example case, the least disturbed background is 542 [j,S/cm
(see Table 6-4). It was estimated at the 25fe centile from the combined EPA-survey and USGS
data set to improve representation of samples from the entire ecoregion. The calculation of the
predicted mean loglO of HC05 (y) is shown in eqs 6-1 and 6-2 and that value is used to estimate
the lower 50% PL using eq 6-3.
6-11

-------
The B-C model is described by the following formula:
Y= Q.651X + 1.075	(6-1)
Where:
Xis the loglO of the ecoregional background SC
Y is the loglO of the predicted HCos
The background for Ecoregion 43 (542 (j,S/cm) is converted to loglO, replacing Xin the
formula with that value and Y is computed (see eq 6.2). The predicted value Y is converted from
loglO to a number that is the modeled HCos for that region. In Ecoregion 43 the mean modeled
HCos is 740 [j,S/cm after rounding to two significant figures.
Log Predicted HCos = (0.657 x 2.73) + 1.075 = 2.87 ^iS/cm	(6-2)
Then
Predicted HCos = 102 87 [j,S/cm = 743 [j,S/cm
6-12

-------
YES
NO
NO
YES-,
YES
NO
YES
YES
YES
Apply mean
B-C modeled
hc05
Apply lower
50% PL B-C
modeled HC05
Apply
new
hc05
Apply XCD
derived HC05
Derive new HC05
by XCD method
>200 paired SC and
biological data?
XCD HC05<
lower 50% PL
B-C modeled
Calculate HC05 with
B-C model and
background SC
XCD HC05 >
mean
B-C modeled
Background <626 pS/cm?
• [HC03] + [S04] > [CI]?
XCD HC05 < mean
modeled and > lower
50% PL modeled
HCnR?
>500 paired SC and biological
data suitable for deriving SC
Derive HC05 by XCD method
and by the B-C model and
background SC
Figure 6-5. Process and decision path case example for Ecoregion 43.
Decision path is highlighted in gray and connected by bold lines. Because there
was no previously derived hazardous concentration at the 5th centile (HCos) of a
taxonomic extirpation concentration distribution (XCD) for Ecoregion 43, and the
background was <626 pS/cm, and there were <200 paired specific conductivity
(SC) measurements, the HCos was calculated with the background-to-criterion
(B-C) model at the lower 50% prediction limit (PL).
6.1.7. Formula for Calculating the Lower 50% Prediction Limit
Because the available paired SC and biological data constitute <200 sites, the HCos was
estimated at the lower 50% PL (see Figure 6-5). The 25th centile of background SC and the
predicted mean HCos of a region and variance of the B-C model is used to calculate the lower
6-13

-------
50% PL (see eq 6-3). In this case example for Ecoregion 43, the background is 542 [j,S/cm and
the mean modeled HCos is 743 [j,S/cm. Both values are converted to loglO (x, y) as shown in
eq 6-4. The prediction interval from the regression line for a mean predicted value y can be
calculated as follows or more conveniently using statistical software thus avoiding rounding
errors:
9 + ta/z.n-zSyJl + ^ + ^p1 = log 10 PL	(6-3)
Symbol
r**\
y
n
a
tn-2
ss
x
o
X
PL
Explanation
LoglO of mean predicted HCos
Number of samples in the model
Alpha error rate for prediction interval
(desired confidence level)
lvalue at specific level (alpha, a) and
degrees of freedom (n - 2) of interval
Residual standard error of prediction
(standard deviation)
Sum of square of x deviation from their
mean, SS = £f=1(Xj — x)2
Mean x values used in the model
generation
x value for a new prediction interval
Upper and lower prediction limits of
mean predicted y
Example from the B-C model
2.87 [j,S/cm, loglO of 743 [j,S/cm
n = 24
50% prediction interval (a = 0.5)
For 50% prediction interval (a = 0.5),
£(1 —0.5)/2,24—2 = 0.686
Sy = 0.11
SS = 4.21
x = 2.15
LoglO 542 [j,S/cm = 2.73
calculated in eq 6-4
Using x° = loglO (542) [j,S/cm and the mean predicted HCos for Ecoregion 43 value,
(y = 2.87, the loglO of 743 (j,S/cm) the lower 50% PL is calculated as follows in eq 6-4. Note,
the upper 50% PL is not calculated but is included in the formula because it may be used to
estimate a CMEC where there are insufficient data to calculate a CMEC using the method
described in Section 3.2.
6-14

-------
loglO (743) ± 0.686 X 0.11Jl+^ + ^Slo(5^ 215)2.	(6-4)
So:
(25 (2.73 - 2.15)2
2.87- 0.686 X 0.11 — +-	—	—
24	4.21
2.87 -0.686 X 0.11 X 1.06
102-79 = 617 ^S/cm
The log of the lower calculated 50% PL is 2.79 which equals 617 [j,S/cm after back
transformation. The lower 50% PL rounded to two significant figures yields a CCC of
620 [j,S/cm.
6.1.8. Example Criterion Maximum Exposure Concentration
At sites meeting the CCC of 620 [j,S/cm, 90% of the SC observations are estimated to
occur below the CMEC (see Section 3.2). The CMEC was derived using the USGS data set
(45,489 samples collected between 1946-2015). This data set was used because it contained
multiple measures of SC within a year whereas the EPA survey data set consisted of single
measurements at each site. Of the 45,489 samples in this ecoregion, there are 41,648 samples in
a July-to-June rotating year representing 1,241 rotation years, 148 unique stations, with at least
1 sample from July to October and one from March to June and at least 6 samples within a
rotation year. Note that inclusion of samples is not contingent on biological data.
Of the 1,241 rotation years (148 unique stations) with multiple SC measurements, the
variability of within station SC slightly differed for streams with different mean SC (see
Figure 6-6). However, the LOWESS indicated that the average variability (residual standard
deviation for a station) was relatively stable (see Figure 6-6); therefore, the entire data set was
used to estimate the standard deviation components of annual mean SC (620 (j.S/cm). The grand
mean and standard deviation of this data set was determined, and the CMEC was calculated. The
example calculation of the CMEC for Ecoregion 43 is shown below using eq 3-2 from
Section 3.2:
6-15

-------
10logl0(620) + 1.28*0.333 = 1?656 ^S/cm
(6-5)
The example CMEC (see Table 6-5) rounded to two significant figures yields a CMEC of
1,700 [j,S/cm for Ecoregion 43. If this level is not exceeded, where the annual geometric mean
SC <620 [j,S/cm, 90% of the observations are expected to be less than the CMEC.
I i i I i i i i i	I ' i I r
200	500 1000 2000	5000
Specific conductivity (^iS/cm)
Figure 6-6. Illustration of within site variability (residual standard deviation
for each station) along the specific conductivity gradient (station mean) in
Ecoregion 43. The x-axis is annual mean specific conductivity. Each dot
represents a station. The fitted line is a locally weighted scatterplot smoothing
(LOWESS, span = 0.75, linear polynomial model).
6-16

-------
Table 6-5. Summary data related to the calculation of the example criterion
maximum exposure concentration (CMEC) for Ecoregion 43
Number of samples July to June prior to biological sampling
41,648
Number of unique stations/rotation years
148/1,241
CCC
620 (j,S/cm
CMEC
1,700 (j,S/cm
6.2. EXAMPLE CRITERION CHARACTERIZATION FOR ECOREGION 43 BASED
ON A BACKGROUND-TO-CRITERION MODEL
The case example for Ecoregion 43 includes an annual geometric mean (i.e., CCC) and a
1-day mean (i.e., CMEC), not to be exceeded more than once in 3 years on average. Both of
these distinct expressions of the example SC criteria would need to be met in order to adequately
protect aquatic life. These values indicate that freshwater animals are protected if the annual
geometric mean SC concentration in flowing waters does not exceed 620 piS/cm and the 1-day
mean does not exceed 1,700 [j,S/cm more than once every 3 years on average. These example
criteria would apply to all flowing freshwaters (ephemeral, intermittent, and perennial streams)
in Ecoregion 43 inclusive of portions of Montana, Wyoming, South Dakota, North Dakota, and
Nebraska. On a site-by-site basis, these example ecoregional criteria apply if the ionic mixture is
dominated by anions of bicarbonate and sulfate and either sodium or calcium cations. For
streams crossing into Ecoregion 43 from ecoregions with either lower or higher background SC,
professional judgment may be needed to assess the potential effect of different ionic composition
or concentration. Professional judgment is recommended when applying to sites with a
catchment area greater than 1,000 km2 (386 mi2) owing to lesser representation in the data set by
this class of stream in the development of the B-C model. On a site by site basis, alternative SC
criteria may be more appropriate if the natural background of a site is shown to be lower or
higher than its regional background SC.
In particular, some streams in Ecoregion 43 may have consistently low SC throughout the
year (Keya Paha Tablelands [43i], Niobara River Breaks [43r], Noncalcareous Foothill
Grasslands [43s], Shield-Smith Valleys [43t], Limy Foothill Grassland [43u], and Pryor-Big
Horn Foothills [43v]). Because all or most of the sampled sites in these Level IV ecoregions
were measured at less than 500 [j,S/cm, a finer resolution (subecoregional) analysis is
6-17

-------
recommended to adequately protect these areas of Ecoregion 43. Specifically, because the more
western parts of the ecoregion provide sources of freshwater and dilution to the rest of the
ecoregion and ecoregions to the east, subdividing the ecoregion according to background SC and
developing different SC criteria may help to protect regional water quality where geophysical
and climatic conditions lead to higher natural SC regimes.
The weight-of-evidence method described in Appendix C could be used to evaluate
subregions or stream classes that may have different background SC in this large ecoregion
where natural background SC may range from <100 [j,S/cm to >1,000 [j,S/cm. For example,
higher criteria may be appropriate for areas such as the Little Missouri Badlands (43b).
6.3. PROTECTION OF FEDERALLY-LISTED SPECIES AND OTHER HIGHLY
VALUED TAXA
Although the example criteria were derived using the macroinvertebrate taxa represented
in the data sets, the available evidence indicates that other taxa in the streams would likely be
protected as well (see Section 2.6 and Appendix G). Hence, no adjustment was made for
unanalyzed taxa. However, on a site-specific basis, the example criterion could be adjusted or
recalculated to protect important species, highly valued aquatic communities, or specially
protected waters.
6-18

-------
7. CASE STUDY IV: EXAMPLE USING THE BACKGROUND TO CRITERION (B-C)
REGRESSION METHOD FOR A REGION WITH LOW CONDUCTIVITY
This section presents an example calculation of an ecoregional CCC using the B-C
method (see Section 3.7.2 and Appendix D). In this example, a CCC for the Cascades Level III
Ecoregion 4 was calculated using SC data from the ecoregion and the B-C method because there
were insufficient paired SC and biological data to accurately estimate the XC95 values for the
XCD method in this ecoregion.
First, the water chemistry data set was screened for ionic composition to ensure samples
were dominated by sulfate and bicarbonate anions and calcium and magnesium cations.
Sampled sites were mapped to determine whether the geographic distribution of sites was
representative of (dispersed throughout) the ecoregion. Minimally affected background SC of
the new ecoregion was estimated at the 25th centile of probability samples (see Section 3.7.1.2;
Cormier and Suter, 2013a). Least disturbed background SC was estimated at the 25th centile of a
combined data set of targeted and probability samples. The CCC was calculated using the
25th centile least disturbed background SC as the independent variable (x) in the B-C regression
model to yield an HC05 (y). Depending on available data and analytical results (see Figure 3-11),
an HC05 may take the form of (1) the value at the mean of the regression line from the B-C
model, (2) the value at the lower 50% PL of the regression line from the B-C model, or (3) an
HC05 derived from a data set based on >200 paired SC and biological samples. In this example
case for Ecoregion 4, the range of SC conditions was narrow with few sites exceeding 200
|iS/cm and only two SC measurements exceeding 1,000 |iS/cm, so any HC05 would be uncertain
using paired SC and biological measurements. Therefore, an HC05 was not calculated using
paired SC and biological measurements and the lower 50% PL was used to develop the example
CCC.
7.1. DATA SET CHARACTERISTICS
7.1.1. Ecoregion Description
The Cascades (Ecoregion 4) is a mountainous ecoregion extending from the central
portion of western Washington into Oregon and, after a separation by the Klamath River, another
separate mountainous area in northern California (U.S. EPA, 2013d). The mountain ranges of
7-1

-------
the Cascades continue northerly into the North Cascades (Ecoregion 77), and south into the
Sierra Nevada (Ecoregion 5). To the east lies the Eastern Cascades Slopes and Foothills
(Ecoregion 9). To the west in Washington state is the Puget Lowlands (Ecoregion 2) and in
Oregon the fertile Willamette Valley (Ecoregion 3). To the west in southern Oregon and
northern California, the Cascades are bounded by the Klamath Mountains/California High North
Coast Range (Ecoregion 78).
Some peaks in Ecoregion 4 are snow-capped or glaciated year round. Both active and
dormant volcanoes are located on the high plateau in the eastern part of the ecoregion. The
highest strato-volcano is Mount Ranier with an elevation of 4,392 m (USGS, 2016). The western
Cascades in Oregon and Washington are dissected by numerous, steep-sided stream valleys
(U.S. EPA, 2013d).
This geologic area is underlain by Cenozoic volcanics that have been affected by alpine
glaciation (U.S. EPA, 2013d). Soils are characterized by frigid temperature regimes and at lower
elevations in the south. Some soils are mesic. Common soils include andisols, formed in
volcanic ash containing high proportions of glass and amorphous colloidal materials, and
inceptisols, nearly like the parent material and having little or no clay, iron, aluminum or organic
matter. The mean annual precipitation ranges from 180 cm and ranges from 115 in the Southern
Cascades to 360 cm on some of the highest peaks of the Cascades Subalpine/Alpine
subecoregion (OWEB 2001, Wilkins et al. 2011). Two of the larger rivers include the Columbia
and Klamath Rivers.
The Cascades have a moist, temperate climate that supports an extensive and highly
productive coniferous forest that is intensively managed for logging. Conifers dominate except
at the highest elevations where there are alpine meadows and rocky alpine zones.
7.1.2. General Data Set Description
Only preexisting water chemistry data were used for this example assessment (see
Table 7-1). EPA-survey, State, and USGS data sets were combined to characterize ion
concentrations and water chemistry and then calculate a provisional CCC. The USGS data set
was also used to calculate the CMEC because this data set had many within-year samples at each
site. The statistical package R, Version 2.12.1 (December 2010), was used for all statistical
analyses (R Development Core Team, 2011).
7-2

-------
Table 7-1. EPA and U.S. Geological Survey (USGS) chemistry data sets
included in this study.
Years indicate the period during which the samples were collected. Western
Environmental Monitoring and Assessment Program (EMAP) survey sites are
included in the count of sites from the National Wadeable Streams Assessment
(NWSA). Specific conductivity was not measured at all sites.
Survey
Years
# of sites
# of samples
EPA probability samples (sample size in parenthesis)
NWS A, NRSA, Region 10 R-EMAP
1995-2009
152
152
Total

152
152
State data from EPA Region 10
Oregon
1990-2014
418
2,511
Washington
1990-2015
121
562
Total (only 90 with SC)

539
3,073
State: subset >_six samples per year, January-December
1990-2015
19
1,111
USGS mixed sampling
USGS: full data set
1958-2016
290
6,258
USGS: subset >_six samples per year, January-December
1959-2014
50
5,019
The B-C regression model was developed using biological data paired with SC data from
24 data sets with waters having ionic mixtures dominated by calcium, magnesium, sulfate and
bicarbonate ions and where background SC did not exceed 626 [j,S/cm. Therefore, the model is
most appropriate for waters with similar ionic characteristics.
Because the B-C regression model was developed with an ionic mixture dominated by
bicarbonate and sulfate (i.e., ([HCO3 ] + [SO42 ]) > [CP] in mg/L), samples dominated by
chloride (i.e., ([HCO3 ] + [SO42 ]) < [Cl~] in mg/L) should be removed from the data set prior to
estimating background SC. In this case, no samples were dominated by chloride ions and so all
samples in the EPA-survey data set were used. The State data set of 539 sites was used to verify
ionic composition. Of 359 samples of the State data set with ionic measurements, all samples
were dominated by sulfate plus bicarbonate. Therefore, none was removed from the final
analysis. All sites in the combined EPA-USGS data set had pH data. Because none were <6 or
>8.7, no sites were removed from the data set.
7-3

-------
7.1.3. EPA-Survey Data Set
Data sources, sampling period, and number of samples used to estimate background SC
in the new area (Ecoregion 4) are provided in Table 7-1. The NRSA 2008-2009 data set
(U.S. EPA, 2013b), NWS A 2004 data set (U.S. EPA, 2006), R-EMAP 1999 data sets (U.S. EPA,
2013c), are based on single random (i.e., probability-based design) samples from June through
September.
EPA-survey data sampling sites within the ecoregion are shown in Figure 7-1. Water
quality parameters collected in Ecoregion 4 are included in Table 7-2. Most of the samples have
reported SC, alkalinity, hardness, sulfate, chloride, bicarbonate, and pH, as well as other water
quality parameters. When necessary, ionic concentrations in milliequivalents (meq/L) were
converted to mg/L [(meq/L) x (ion MW)/(ionic charge)] (Hem, 1985).
Table 7-2. Summary of data for the example case for Cascades, Level III
Ecoregion 4, from EPA-survey samples.
Calculated means are geometric except for pH values and ion ratios.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
152
1.56
33.9
44.9
44.8
61.8
205.0
pH (SU)
144
6.17
7.3
7.5
7.5
7.7
9.0
Ca2+ (mg/L)
70
0.04
3.1
4.8
4.7
8.1
22.7
Mg2+(mg/L)
70
0.03
0.7
1.3
1.2
1.9
OO
00
Na+ (mg/L)
70
0.08
2.1
2.8
2.7
4.0
10.4
K+ (mg/L)
70
0.01
0.2
0.5
0.5
0.8
3.2
HCO3 (mg/L)
46
0.46
18.3
26.4
25.4
40.0
132.2
S042 (mg/L)
121
0.07
0.4
1.0
1.0
2.3
34.3
cr (mg/L)
136
0.14
0.6
0.9
0.8
1.1
12.2
a hco3 + so42 /cr
46
3.88
24.5
32.3
50.4
58.4
211.1
aRatio of mg/L HCO3 + SO42 /CI greater than 1 indicates the mixture is dominated by HCO3 + SO42 .
All samples were collected from first-through fourth-order streams as part of a
probability-based design intended to estimate proportions of parameters for various stream
classes. The probability-design sampling weights for stream order were not used in the
7-4

-------
characterization. Analysis of water chemistry samples followed procedural and QA/QC
protocols of EPA andEMAP (U.S. EPA, 2001, 1998b, 1994, 1987), Wadeable Streams
Assessment (U.S. EPA 2006, 2004a, b), NRSA (U.S. EPA, 2009), and NAPAP (Drouse et al.,
1986; U.S. EPA, 1987).
Figure 7-1. Sampling sites in the EPA survey data set that were used to
estimate minimally affected background in Ecoregion 4.
For the purpose of demonstrating the background-to-criterion approach, Level III
Ecoregion 4, is shaded gray. A total of 152 sites with specific conductivity
measurements are depicted. Sampling locations are color coded by site-specific
conductivity range: green diamonds <30 [j,S/cm, yellow squares 30-100 [j,S/cm,
and red triangles >100 [j,S/cm. Geodetic reference system = North American
Datum (NAD83).
WA
OR
Specific Conductivity
C nS/cm)
° <30
° 30-100
A >=100
7-5

-------
7.1.4. State Data Set (EPA Region 10)
The State data set includes data from the Oregon DEQ 1990-2014 data set (Oregon
DEQ, 2009) and the Washington Department of Ecology (WDE) 1990-2015 data set (WDE,
2014). Monthly sampling was performed throughout the year in many sites.
State sampling sites within the ecoregion are shown in Figure 7-2. Water quality
parameters collected in Ecoregion 4 are included in Table 7-3. Most of the samples have
reported SC, sulfate, chloride, and pH.
WA
mo
100
200 km
OR
CA
Specific Conductivity
(^iS/cm)
° 30-100
* >=100
Figure 7-2. Sampling sites in State data set that were used to estimate
minimally affected background in Ecoregion 4.
For the purpose of demonstrating the background-to-criterion approach, Level III
Ecoregion 4 is shaded gray. A total of 539 sampling sites are in the state data set,
but only 95 sites with conductivity measurements are shown here. Sampling
locations are color coded by site-specific conductivity range: green diamonds
<30 [j,S/cm, yellow squares 30-100 [j,S/cm, and red triangles >100 [j,S/cm.
Geodetic reference system = North American Datum (NAD83).
7-6

-------
Table 7-3. Summary of data for the example case for Cascades, Level III
Ecoregion 4 from State Data from Oregon and Washington.
Calculated means are geometric except for pHs and ion ratios.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
95
6.8
24.0
32.4
31.9
39.4
126
Ca2+ (mg/L)
78
1.0
2.3
3.0
3.7
4.3
1,100
HCO3 (mg/L)
300
7.7
19.9
25.8
27.5
38.0
3,966
S042 (mg/L)
252
0.1
0.3
0.6
0.8
1.7
36.2
cr (mg/L)
315
0.3
0.7
1.0
1.1
1.5
250.0
3hco3 + so42 /cr
203
3.3
18.0
28.2
37.6
46.0
225.7
aratio of mg/L HCO3 + SO42 /CI greater than 1 indicates the mixture is dominated by HCO3 + SO42 .
7.1.5. U.S. Geological Survey (USGS) Data Set
The USGS survey data set used in this example is composed of 6,258 water quality
samples, from 290 stations within Ecoregion 4 (see Table 7-1; Figure 7-3). Some stations were
sampled only once while others were sampled as many as 641 times. The data were collected
between 1958 and 2016 during all seasons. Water quality parameters collected in Ecoregion 4
are included in Table 7-4. Most of the samples have reported SC, alkalinity, hardness, chloride,
bicarbonate, and pH, as well as other water quality parameters. Analysis of water chemistry
samples followed procedural and QA/QC protocols for USGS data sets (Mueller et al. 1997).
7-7

-------
Figure 7-3. Sampling sites in the U.S. Geological Survey (USGS) data set in
Ecoregion 4.
Ecoregion 4 is shaded gray. Geometric mean specific conductivity at
290 sampling locations is color-coded: green diamonds <30 [j,S/cm, yellow
squares 30-100 [j,S/cm, and red triangles >100 [j,S/cm. Geodetic reference
system = North American Datum (NAD83).
Specific Conductivity
(nS/cm)
•	<30
° 30-100
*	>=100
7-8

-------
Table 7-4. Summary of data for the example case for Cascades, Level III
Ecoregion 4 from the U.S. Geological Survey (USGS).
Calculated means are geometric except for pH.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
282
3.50
38
53.9
53.6
78.3
370
pH (SU)
274
6.2
7.2
7.5
7.4
7.7
8.7
Ca2+ (mg/L)
5
1.6
1.7
2.1
2.9
5.6
15.7
Mg2+ (mg/L)
158
0.10
0.78
1.39
1.31
2.43
9.08
Na+ (mg/L)
5
1.2
1.2
1.3
1.7
2.5
5.6
K+ (mg/L)
5
0.1
0.2
0.2
0.3
0.5
1.3
HCO3 (mg/L)
57
3.0
29.8
35.0
37.0
51.3
122.0
All samples were collected from first-through fourth-order streams for various research
purposes, and those targeted sampling designs may emphasize areas with increased
anthropogenic disturbance. In this respect, the data may skew the background estimates.
Background SC was not estimated from this data set alone because sampling stations were not
randomly assigned.
Because the USGS and State data sets contained multiple measurements in a year from
sampling locations, the data sets were used to estimate a CMEC and explore the variability of SC
patterns within the region. Therefore, a second data set was selected by excluding stations with
fewer than 6 SC measurements throughout the year. A minimum of one sample during the
spring (March to June) and one in the summer (July to October) were also required so that at
least one lower and one higher SC sample were included in the data set. The second USGS data
set included 5,019 samples from 50 stations, while the second State data set include
1,111 samples from 19 stations. The USGS and State data sets were combined was used to
calculate the variance near the CCC for the CMEC calculation.
7.1.6. Modeled Mean Base Flow Background Specific Conductivity (SC)
Predicted mean natural base-flow specific conductivity in catchments of Cascades, was
also considered for comparison purposes (Olson and Hawkins, 2012). The stream length
weighted, mean natural SC from the modeled base flow were calculated for the SC at base flow
7-9

-------
for each ecoregion (see Appendix D). Figure 7-4 shows the predicted SC with 300 m resolution
in order to emphasize the general trends across Ecoregion 4.
N
w
Specific
conductivity (fiS/cm)
n- 300
WA
- 250
OR
-200
- 150
- 100
CA
150
300 km
Alters Equal Area Conic USGS
Figure 7-4. Predicted mean natural base-flow specific conductivity in
catchments of the Cascades, Ecoregion 4, using the Olson-FIawkins model.
Albers projection used for mapping.
7-10

-------
7.1.7. Characterization of Ionic Matrices
7.1.7.1.	EPA-Survey Data Set Ionic Characteristics
A summary of water quality ionic constituents including major ionic constituents from
EPA-survey data for Ecoregion 4 is provided in Table 7-2. Centiles were calculated using each
sample observation because most measurements were single grab samples. There were no
chloride-dominated sites where [CP] > ([HCO3 ] + [SO42 ]) in mg/L in the EPA-survey data
(N= 152); therefore, no sites were excluded from the data sets. Calcium plus magnesium were
the dominant cations ([Ca2+] + [Mg2+]) > ([Na+] + [K+]).
7.1.7.2.	State Data Set Ionic Characteristics
A summary of water quality ionic constituents including major ionic constituents from
State data for Ecoregion 4 is provided in Table 7-3. Centiles were calculated using each sample
observation because most measurements were single grab samples. There were no
chloride-dominated sites (where [CP] > ([HCO3 ] + [SO42]) in the state data (N= 539);
therefore, no sites were excluded from the data sets. Likewise, calcium plus magnesium were
the dominant cations, ([Ca2+] + [Mg2+]) > ([Na+] + [K+]).
7.1.7.3.	U.S. Geological Survey (USGS) Data Set Ionic Characteristics
Table 7-4 summarizes the water quality parameters including major ionic constituents for
Ecoregion 4 in the USGS data set. Unlike the EPA-survey data set, the USGS data set are
targeted sites of interest rather than probability samples. Also, in some cases, there are multiple
measurements from the same site, and other sites are auto-correlated with downstream sampling
locations. Therefore, the distribution of sites in this data set is not necessarily representative of
Ecoregion 4 in its entirety; however, the data set can be used to define the overall pattern of SC
for the ecoregion because it contains samples in areas not represented in the EPA-survey data set.
Centiles were calculated using the geometric mean of site measurements (except pH and the
ionic ratio) and were qualitatively compared to the probability-based EPA-survey data set. No
sulfate measurements were found so ([HCO3 ] + [SO42 ])/[Cl ] ratios were not determined for
this data set. Calcium plus magnesium were the dominant cations
([Ca2+] + [Mg2+]) > ([Na+] + [K+]).
7-11

-------
Two samples from the USGS data set were in the 1,000 [j,S/cm range and were associated
with salt springs at Lake Paulina and Longmire Meadow mineral springs and were removed from
the data set (Ingebritsen et al., 2014). Six sites (421-1,030 (j,S/cm) were sampled on the flanks
of Mount St. Helen after the 1980 eruption; and, these were removed. These eight sites represent
a small proportion of the data set, and the example criterion would not apply to these or similar
areas with naturally higher background SC. The background SC was 39.5 [j,S/cm with these
eight sites included and 38.2 [j,S/cm with them removed.
7.1.8. Comparison of Background Specific Conductivity (SC) Estimates
SC data from the three data sets, the combined data, and the Olson-Hawkins base flow
model are summarized in Figure 7-5. The stream length weighted average predicted mean base
flow SC from the Olson-Hawkins model in Ecoregion 4 is 65.7 [j,S/cm. The 25th centile of the
EPA-survey data set is 33.9 [j,S/cm, of the State data set is 24 [j,S/cm, and of the USGS SC data
set is 39.5 [j,S/cm. All three are less than the mean base flow SC (65.7 (j,S/cm), so these
background SC are characterized as minimally affected (see Figure 7-5 and Tables 7-2, 7-3, and
7-4). When data are combined, the 25th centile (-33 (j,S/cm) is still lower than the predicted
mean base flow of 65.7 [j,S/cm (see Table 7-5, Figure 7-5). Therefore, this background SC
estimated from the combined State-EPA-USGS data set also represents minimally affected
background SC.
The 25th centile of the combined data set was used to calculate the HCos following the
decision tree described Section 3.7.2 and Figure 3-11.
7-12

-------
1000 -
500 -
200 -
100 -
50
20 -
10 -
	1	1	1—
Baseflow Combined EPA
	1	
USGS
State
Figure 7-5. Box plots of specific conductivity (SC) distributions for
EPA-survey, State, U.S. Geological Survey (USGS), and combined data sets,
and the predicted base flow SC for all stream segments.
The State data set captures a slightly narrower range of values possibly owing to
the targeted sampling. The USGS samples greater than 400 |iS/cm include some
samples from mineral springs and Mt. St. Helen ash flows which have been
removed from the combined data set. The 25th centiles of the observed data sets
are fairly similar to the 25th centile of predicted base-flow.
7-13

-------
Table 7-5. Summary of data for the example case for Cascades, Level III
Ecoregion 4 from the combined data set.
Calculated means for sampled sites are geometric for all variables except for pH
and ion ratio.
Parameter
N
Minimum
25th
50th
Mean
75th
Maximum
SC (nS/cm)
529
1.56
32.7
46.4
46.4
66
370
pH (SU)
418
6.2
7.2
7.5
7.4
7.7
9.0
Ca2+ (mg/L)
153
0.04
2.51
3.59
4.06
6.36
1,100
Mg2+ (mg/L)
228
0.03
0.77
1.34
1.29
2.20
9.08
Na+ (mg/L)
75
0.08
2.03
2.76
2.59
3.96
10.45
K+ (mg/L)
75
0.01
0.22
0.49
0.44
0.77
3.17
HCO3 (mg/L)
403
0.46
20.4
28.0
28.4
40.2
3,966
S042 (mg/L)
373
0.07
0.34
0.78
0.85
2.17
36.2
cr (mg/L)
451
0.14
0.63
1.00
1.00
1.40
250
hco3 + so42 /cr
249
3.29
18.3
28.6
40
48.5
226
"ratio in mg/L HCO3 + SO42 /CI greater than 1 indicates the mixture is dominated by HCO3 + SO42 .
7.2. CALCULATION OF THE CRITERION CONTINUOUS CONCENTRATION (CCC)
7.2.1. Calculation of the Ecoregion 4 mean Hazardous Concentration (HC05) from
Background
Because paired SC and biological data are available for <200 sites, the HC05 was estimated
at the lower 50% PL of the B-C model (see Figure 7-6). The 25th centile of SC from the
combined data set of Ecoregion 4 was used to identify the lower 50% PL using eqs 3-4 and 3-5.
The B-C model development is described in Appendix D. The x-variable is the background SC
in Ecoregion 4 which was loglO transformed. The calculated .y-value is the predicted mean
loglO HC05. In this example case, the minimally affected background is 33 [j,S/cm (see
Table 7-5). It was estimated at the 25th centile from the combined data set. The calculation of
the predicted mean loglO of HC05 (y) is shown in eqs 7-1 and 7-2 and that value is used to
estimate the lower 50% PL using eq 7-3.
7-14

-------
YES
NO
NO
YES-,
YES
NO
YES
YES
YES
Apply mean
B-C modeled
hc05
Apply lower
50% PL B-C
modeled HC05
Apply
new
hc05
Apply XCD
derived HC05
Derive new HC05
by XCD method
>200 paired SC and
biological data?
XCD HC05<
lower 50% PL
B-C modeled
Calculate HC05 with
B-C model and
background SC
XCD HC05 >
mean
B-C modeled
Background <626 pS/cm?
• [HC03] + [S04] > [CI]?
XCD HC05 < mean
modeled and > lower
50% PL modeled
HCnR?
>500 paired SC and biological
data suitable for deriving SC
Derive HC05 by XCD method
and by the B-C model and
background SC
Figure 7-6. Process and decision path case example for Ecoregion 4.
Decision path is highlighted in gray and connected by bold lines. Because there
was no previously derived hazardous concentration at the 5th centile (HCos) of a
taxonomic extirpation concentration distribution (XCD) for Ecoregion 4, and the
background was <33 pS/cm, and there were <200 paired specific conductivity
(SC) measurements, the HCos was calculated with the background-to-criterion
(B-C) model at the lower 50% prediction limit (PL).
7-15

-------
The B-C model is described by the following formula:
Y= Q.651X + 1.075
(7-1)
Where:
Xis the loglO of the ecoregional background SC
Y is the loglO of the predicted HCos
The background for Ecoregion 4 (33 (j,S/cm) is converted to loglO, replacing Xin the
formula with that value and Y is computed (see eq 7-2). The predicted value Y is converted from
loglO to a number that is the mean modeled HCos for that region. In Ecoregion 4 the mean
modeled HCos is 118 [j,S/cm after rounding to two significant figures.
7.2.2. Calculation of the Lower 50% Prediction Limit
Because the available paired SC and biological data constitute <200 sites, the CCC was
estimated at the lower 50% PL of the HCos (see Figure 7-6). The 25th centile of background SC
and the predicted mean HCos of a region and variance of the B-C model is used to calculate the
lower 50% PL (see eq 7-3). In this case example for Ecoregion 4, the background is 33 [j,S/cm
and the mean modeled HCos is 118 [j,S/cm. Both values are converted to loglO (x, y) as shown in
eq 7-4. The prediction interval from the regression line for a mean predicted value y can be
calculated as follows or more conveniently using statistical software thus avoiding rounding
errors:
Log Predicted HCos = (0.657 x 1.518 jiS/cm) + 1.075 = 2.072 ^iS/cm
(7-2)
Then
Predicted HCos = 102072[j,S/cm = 118 [j,S/cm
(7-3)
7-16

-------
Symbol	Explanation	Example from the B-C model
y	LoglO of mean predicted HCos 2.072 [j,S/cm, loglO of 118 [j,S/cm
n	Number of samples in the model n = 24
a	Alpha error rate for prediction	50% prediction interval (a = 0.5)
interval (desired confidence level)
tn~2 lvalue at specific level (alpha, a) For 50% prediction interval (a = 0.5),
and degrees of freedom (n - 2) of t(i-o.5)/2,24-2 = 0.686
interval
Sy	Residual standard error of prediction Sy = 0,1 1
(standard deviation)
SS	Sum of square of x deviation from SS = 4.21
their mean, SS = £f=i(xi — x)2
x	Mean x values used in the model x = 2.15
generation
x	x value for a new prediction interval LoglO 118 [j,S/cm = 2.072
PL Upper and lower prediction limits of calculated in eq 6-4
mean predicted y
Using x° = loglO (118) [j,S/cm and the mean predicted HCos for Ecoregion 4 value,
(y = 2.072, the loglO of 118 (j,S/cm) the lower 50% PL is calculated as follows in eq 7-4. Note,
the upper 50% PL is not calculated but is included in the formula because it may be used to
estimate a CMEC where there are insufficient data to calculate a CMEC using the method
described in Section 3.2.
log 10(118) ± 0.686 x 0.11 ll + ^ + (l03lo(1^ 215)2	(7.4)
So:
25 (2.072 - 2.15)2
2.072 - 0.686 X 0.11 —	^
24	4.21
2.072 -0.686 X 0.11 X 1.02
LoglO1"7 = 98 [iS/cm
7-17

-------
The log of the lower calculated 50% PL is loglO 1.997 which equals 98 [j,S/cm after back
transformation and rounding to two significant figures yields a CCC of 98 [j,S/cm.
7.2.3. Example Criterion Maximum Exposure Concentration
At sites meeting the CCC of 98 [j,S/cm, 90% of the SC observations are estimated to
occur below the CMEC (see Section 3.2). The CMEC was derived using a combined USGS
(6,258 samples collected between 1959-2016) and State data sets (3,073 samples collected
between 1990-2015). These data sets were used because they contained multiple measures of
SC within a year whereas the EPA survey data set consisted of single measurements at each site.
Of the 9,331 samples in this ecoregion, there are 6,130 samples in a year representing 312 station
years, 69 unique stations, with at least 1 sample from March to June and one from July to
October and at least 6 samples within a year. Note that inclusion of samples is not contingent on
biological data.
Of the 312 station years (69 unique stations) with multiple SC measurements, the
variability of within station SC was slightly differed for streams with different mean SC (see
Figure 7-7). However, the LOWESS and confidence bounds for any detectable change points
indicated that the average variability (residual standard deviation for a station) was relatively
stable generally between 0.05 and 0.1 (see Figure 7-7); therefore, the entire data set was used to
estimate the standard deviation components of the annual mean SC (98 (j,S/cm). The proposed
CCC and standard deviation of this data set was determined, and the CMEC was calculated. The
example calculation of the CMEC for Ecoregion 4 is shown below (see eq 7-5) using eq 3-2
from Section 3.2:
10loglO(98) + 1.28*0.234 = 1% ^s/cm	(?_5)
7-18

-------
0.25 -
0.20 -
sz
o
-4—1
ro
£ 0.15 -
"O
"O
ro
"O
ra 0.10 -
-I—«
CO
0.05 -
Figure 7-7. Illustration of within site variability (residual standard deviation
for each station) along the specific conductivity gradient (station means) in
Ecoregion 4.
The x-axis is log annual mean specific conductivity. Each dot represents a
station. The fitted line is a locally weighted scatterplot smoothing spline
(LOWESS, span = 0.75, linear polynomial model).
The example CMEC (see Table 7-6) rounded to two significant figures yields a CMEC of
200 pS/cm for Ecoregion 4. At this level, where the annual geometric mean SC <98 pS/cm, 90%
of the observations are expected to be less than the CMEC.
Table 7-6. Summary data related to the calculation of the example criterion
maximum exposure concentration (CMEC) for Ecoregion 4
Number of samples July to June prior to biological sampling
6,130
Number of unique stations/rotation year
69/312
ccc
98 (j,S/cm
CMEC
200 (j,S/cm
o o
S> O'^oo ~
1=0
fo
10
50
100
Specific conductivity ( ^S/cm)
7-19

-------
7.3. EXAMPLE CRITERION CHARACTERIZATION FOR ECOREGION 4 BASED ON
A BACKGROUND-TO-CRITERION MODEL
The case example for Ecoregion 4 includes an annual geometric mean (i.e., CCC) and a
1-day mean (i.e., CMEC), not to be exceeded more than once in 3 years on average. Both of
these distinct expressions of the example SC criteria would need to be met in order to adequately
protect aquatic life. These values indicate that freshwater animals are protected if the annual
geometric mean SC concentration in flowing waters does not exceed 98 piS/cm and the 1-day
mean does not exceed 200 [j,S/cm more than once every 3 years on average. These example
criteria would apply to all flowing freshwaters (ephemeral, intermittent, and perennial streams)
in Ecoregion 4 inclusive of portions of Washington, Oregon, and California. On a site-by-site
basis, these example ecoregional criteria apply if the ionic mixture is dominated by anions of
bicarbonate and sulfate and cations calcium and magnesium. For streams crossing into
Ecoregion 4 from ecoregions with either lower or higher background SC, professional judgment
may be needed to assess the potential effect of different ionic composition or concentration.
Professional judgment is recommended when applying to sites with a catchment area greater
than 1,000 km2 (386 mi2) owing to lesser representation in the data set by this class of stream in
the development of the B-C model. On a site by site basis, alternative SC criteria may be more
appropriate if the natural background of a site is shown to be lower or higher than its regional
background SC.
The Cascades ecoregion has less sources of ionic inputs and the igneous geology leads to
very low stream SC (background SC of 33 (j,S/cm), which represents minimally affected
conditions with respect to SC. Reference sites were not identified in the data sets so a
comparison with the 75th centile SC in any data set was not possible. About 88% of the sampled
sites (537) meet the CCC and more than 99% of all samples (7,855) meet the CMEC calculated
for this example. Owing to the very low conductivity, there is very little difference between the
lower 50% prediction interval or the mean modeled HCos, 98 versus 118 [j,S/cm, respectively.
Two samples from the USGS data set were in the 1,000 [j,S/cm range and were associated with
salt springs at Lake Paulina and Longmire Meadow mineral springs and were removed from the
data set (Ingebritsen et al., 2014). Six sites (421-1,030 (j,S/cm) were sampled on the flanks of
Mount St. Helen after the 1980 eruption, and they were also removed before the analysis. The
example criterion would not apply to these areas with naturally higher background SC. The
7-20

-------
weight-of-evidence method described in Appendix C could be used to evaluate subregions or
stream classes that may have different background SC in this large ecoregion. In particular, the
isolated area in Northern California may have a naturally higher background SC based on the
USGS measurements (see Figure 7-3) and the mean predicted baseflow (see Figure 7-4). Also,
potential unique sources of salt such as fumaroles and salt springs may naturally raise SC.
7.4. PROTECTION OF FEDERALLY-LISTED SPECIES AND OTHER HIGHLY
VALUED TAXA
Although the example criteria were derived using XC95 values for the macroinvertebrate
taxa represented in the data sets used to develop the B-C model, the available evidence indicates
that other taxa in the streams would likely be protected as well (see Section 2.6 and
Appendix G). Hence, no adjustment was made for unanalyzed taxa. However, on a site-specific
basis, the example criterion could be adjusted or recalculated to protect important species, highly
valued aquatic communities, or specially protected waters.
7-21

-------
8. REFERENCES
Allan, JD. (1981) Determinants of diet of brook trout (Salvelinus fontinalis) in a mountain stream. Can J Fish Aqua
Sci 38:184-92.
Andersen, CB; Lewis, GP; Sargent, KA. (2004) Influence of wastewater treatment effluent on concentrations and
fluxes of solutes in the Bush River South Carolina, during extreme drought conditions. Environ Geosci
11(1):28—41.
Anning, DW; Bauch, NJ; Gerner, SJ; Flynn, ME; Hamlin, SN; Moore, SJ; Schaefer, DH; Anderholm, SK; Spangler,
LE. (2007) Dissolved solids in basin-fill aquifers and streams in the Southwestern United States. U.S.G.S. Scientific
Investigations Report 2006-5315, version 1.1. Updated 2010. Denver, CO: USGS Information Services. Available
online at http://pubs.usgs.gov/sir/2006/5315/.
Anning, DW; Flynn, ME. (2014) Dissolved-solids sources, loads, yields, and concentrations in streams of the
conterminous United States: U. S. Geological Survey Scientific Investigations Report 2014-5012, 101 p. Available
online at http://dx.doi.org/10.3133/sir20145Q12.
APHA (American Public Health Association) (1992) Standard methods for the examination of water and
wastewater. 18th edition. Washington, DC: American Public Health Association.
Barbour, MT; Gerritsen, J; Snyder, BD; Stribling, JB. (1999) Rapid bioassessment protocols for use in streams and
wadeable rivers: periphyton, benthic macroinvertebrates, and fish. 2nd ed. EPA/84l/B-99/002. Washington, DC:
U.S. Environmental Protection Agency, Office of Water. Available online at
http://www.epa. gov/owow/monitoring/rbp/wp6 lpdf/rbp.pdf.
Barlow, PM. (2003) Ground water in freshwater-saltwater environments of the Atlantic Coast. Circular 1262.
Reston, VA: U.S. Geological Survey. Available online at: http://pubs.usgs.gov/circ/2003/circl262/pdf/circl262.pdf.
Barlow, PM; Reichard, EG. (2010) Saltwater intrusion in coastal regions of North America. Hydrogeo J
18:247-260.
Belli, NM; Faleiros, RO; Firmino, KCS; Masui, DC; Leone, FA; McNamara, JC; Furriel, RPM. (2009) Na,K-
ATPase activity and epithelial interfaces of the freshwater shrimp Macrobrachium amazonicum (Decapoda,
Palaemonidae). Comp Biochem Physiol A Mol Integr Physiol 152(3):431-439.
Bernhardt, ES; Lutz, BJ; King, R; Fay, J; Carter, CA; Helton, AM; Amos, J; Campagna, D. (2012) How many
mountains can we mine? Examining the regional impact of surface mining on freshwater ecosystems of the Central
Appalachians. Environ Sci Technol. 46:8115-8122.
Bernhardt, ES; Palmer, MA. (2011) The environmental costs of mountaintop mining valley fill operations for
aquatic ecosystems of the Central Appalachians. Ann NY Acad Sci 1223:39-57Bernot, MJ; Tank, JL; Royer, TV;
David, MB. (2006) Nutrient uptake in streams draining agricultural catchments of the Midwestern United States.
Freshw Biol 51(3):499-509.
Bernot, MJ; Tank, JL; Royer, TV; David, MB. (2006) Nutrient uptake in streams draining agricultural catchments of
the Midwestern United States. Freshw Biol 51(3):499-509.
Berra, TM. (2007) Freshwater fish distribution. Chicago: The University of Chicago Press.
Blasius, BJ; Merritt, RW. (2002) Field and laboratory investigations on the effects of road salt (NaCl) on stream
macroinvertebrate communities. Environ Pollut 120(2):219-231.
Boelter, AM; Lamming, FN; Farag, AM; Bergman, HL. (1992) Environmental effects of saline oil-field discharges
on surface waters. Environ Toxicol Chem 11(8): 1187-1195.
8-1

-------
Bos, DG; Cumming, BF; Watters, CE; Smol, JP. (1996) The relationship between zooplankton, SC and lake-water
ionic composition in 111 lakes from the Interior Plateau of British Columbia, Canada. Int J Salt Lake Res 5(1):
1-15.
Bradley, TJ. (2009) Chapter 7. Hyper-regulators: life in fresh water. In: Animal osmoregulation. New York: Oxford
University Press; pp 149-151.
Bringolf, RB; Cope, WG; Eads, CB; Lazaro, PR; Barnhart, MC; Shea, D. (2007) Acute and chronic toxicity of
technical-grade pesticides to glochidia and juveniles of freshwater mussels (Unionidae). Environ Toxicol Chem
26(10):2086-2093.
Brinck, EL; Drever, JI; Frost, CD. (2008) The geochemical evolution of water coproduced with coalbed natural gas
in the Powder River Basin, Wyoming. Environ Geosci 15(4): 153—171.
Bringolf, RB; Cope, WG; Eads, CB; Lazaro, PR; Barnhart, MC; Shea, D. (2007) Acute and chronic toxicity of
technical-grade pesticides to glochidia and juveniles of freshwater mussels (Unionidae). Environ Toxicol Chem
26(10):2086-2093.
Brittain, JE. (1982) Biology of mayflies. Ann Rev Entomol. 27:119-147.
Brittain, JE; Eikeland, TJ. (1988) Invertebrate drift - a review. Hydrobiologia 166:77-93.
Brock, TCM; Roessink, I; Belgers, JDN; Bransen, F; Maund, SJ. (2009) Impact of a benzoyl urea insecticide on
aquatic macroinvertebrates in ditch mesocosms with and without nonsprayed sections. Environ Toxicol Chem
28(10):2191-2205.
Brown, RC; Davis, TP. (2002) Calcium Modulation of Adherens and Tight Junction Function A Potential
Mechanism for Blood-Brain Barrier Disruption After Stroke. Stroke. 33(6): 1706-11.
Bryant, G; McPhilliamy, S; Childers, H. (2002) A survey of the water quality of stream in the primary region of
mountaintop/valley fill coal mining, October 1999 to January 2001. In: Draft programmatic environmental impact
statement on mountaintop mining/valley fills in Appalachia—2003. Appendix D. Philadelphia, PA: U.S.
Environmental Protection Agency, Region 3. Available online at http://www.cet.edu/pdf/mtmvfchemistrv.pdf.
Burton, TM. (1976) An analysis of the feeding ecology of the salamanders (Amphibia, Urodela) of the Hubbard
Brook Experimental Forest, New Hampshire. J Herpetol 10:187-204.
Caneda-Argiielles, M; Kefford, BJ; Piscart, C; Prat, N; Schafer, RB; Schulz, C. (2013) Saliniation of rivers: an
urgent ecological issue. Environ Pollut 173:157-167.
Caquet, T; Hanson, ML; Roucaute, M; Graham, DW; Lagadic, L. (2007) Influence of isolation on the recovery of
pool mesocosms from the application of an insecticide II. Benthic macroinvertebrate responses. Environ Toxicol and
Chem 26(6): 1280-1290.
Cardinale, BJ. (2011) Biodiversity improves water quality through niche partitioning. Nature 472:86-89.
Carey, RO; Migliaccio, KW. (2009) Contribution of wastewater treatment plant effluents to nutrient dynamics in
aquatic systems: A review. Environ Manage 44(2):205-217.
Carter, JL; Resh, VH; Hannaford, J; Myers, MJ. (2006) Macroinvertebrates as biotic indicators of environmental
quality. In: Hauer, FR; Lamberti, GA; eds. Methods in stream ecology, 2nd edition. Burlington, MA: Academic
Press, pp. 805-831.
Clare, EL; Barber, BR; Sweeney, BW; Herbert, PD; Fenton, MB. (2011) Eating local: Influence of habitat on the
diet of little brown bats (Myotis lucifugus). Mol Ecol 20:1772-1780.
8-2

-------
Clark, ML; Miller, KA; Brooks, MH. (2001) U.S. Geological Survey monitoring of Powder River Basin
stream-water quantity and quality. Cheyenne, WY: U.S. Geological Survey, U.S. Department of the Interior,
Cheyenne, WY. Available online at http://pubs.usgs.gov/wri/wriO 14279/.
Clements, WH; Kotalik, C; Cadmus, P. (2014) Effects of major ions onbenthic communities: an experimental test
of the proposed U.S. EPA aquatic life benchmark for conductivity. In: Joint Aquatic Sciences Meeting: Bridging
genes to ecosystems: aquatic science at a time of rapid changes, Portland, Oregon, May 18-23, 2014. Available
online at: http://www.sgmeet.com/iasm2014/viewabstract.asp?AbstractID= 14814.
Clements, WH; Kotalik, C. (2016) Effects of major ions on natural benthic communities: an experimental
assessment of the U.S. EPA aquatic life benchmark for conductivity. Freshw. Sci. 35(1):(126-138).). Available
online at: http://www.iournals.uchicago.edu/doi/pdliplus/10.1086/685085.
Clifford, HF. (1982) Life cycles of mayflies (Ephemeroptera), with special reference to voltinism. Quaest Entomol
18:15-90.
Coffey, DB; Cormier, SM; Harwood, J. (2014) Using field-based species sensitivity distributions to infer multiple
causes. HERA 20(2):402-432.
Colwell, RK; Rangel, TF. (2009) Hutchinson's duality: The once and future niche. Proc Natl Acad Sci U.S.A.
106:19651-19658.
Cormier, SM; Coffey, DB; Griffith, M. (2012) Letter to the Editor in Chief concerning the article "Status of fish and
macroinvertebrate communities in a watershed experiencing high rates of fossil fuel extraction: Tenmile Creek, a
Major Monongahela River tributary" by Kimmel and Argent, 2012. Water Air Soil Pollut 223:4659-4662.
Cormier, SM; Suter, GW. (2013a) A method for deriving water-quality benchmarks using field data. Environ
Toxicol Chem 32(2):255-262.
Cormier, SM; Suter, GW. (2013b) A method for assessing causation of field exposure-response relationships.
Environ Toxicol Chem 32(2):272-276.
Cormier, SM; Suter, GW, II; Norton, SB. (2010) Causal characteristics for ecoepidemiology. HERA 16(1):53—73.
Cormier, SM; Suter, GW; Zheng, L. (2013a) Derivation of a benchmark for freshwater ionic strength. Environ
Toxicol Chem 32(2):263-271.
Cormier, SM; Suter, GW; Zheng, L; Pond, GJ. (2013b) Assessing causation of the extirpation of stream
macroinvertebrates by a mixture of ions. Environ Toxicol Chem 32(2):277-287.
Cormier, SM; Wilkes, SP; Zheng, L. (2013c) Relationship of land use and elevated ionic strength in Appalachian
watersheds. Environ Toxicol Chem 32(2):296-303.
Coward, K; Bromage, NR; Hibbitt, O; Parrington, J. (2002) Gamete physiology, fertilization and egg activation in
teleost fish. Rev Fish Biol Fish 12:33-58.
Crossland, NO; Mitchell, GC; Bennett, D; Maxted, J. (1991) An outdoor artificial stream system designed for
ecotoxicological studies. Ecotox Environ Saf 22:175-183.
Dahm, KG; Guerra, KL; Xu, P; Drewes, JE. (2011) Composite geochemical database for coalbed methane produced
water quality in the Rocky Mountain region. Environ Sci Technol 45(18):7655-7663.
Datry, T. (2012) Benthic and hyporheic invertebrate assemblages along a flow intermittence gradient: effects of
duration of dry events. Freshw Biol (57)563-574.
De Jong, GD; Canton, SP. (2013) Presence of long-lived taxa and hydrologic permanence. J Freshw Ecol
28(2):277-282.
8-3

-------
del Rosario RB, Resh VH (2000) Invertebrates in intermittent and perennial streams: is the hyporheic zone a refuge
from drying? J N Am Benthol Soc 19:680-696.
Delucchi, CM. 1988. Comparison of community structure among streams with different temporal flow regimes.
Can J Zool 66: 579-586.
Doeg, TJ; Milledge, GA. (1991) Effect of experimentally increasing concentration of suspended sediment on
macroinvertebrate drift. Aust J Mare Freshw Res 42(5):519-526.
Drouse, SK; Hillman, DC; Creelman, LW; Potter, JF; Simon, SJ. (1986) National surface water survey, Eastern
Lake survey - Phase I, quality assurance plan. EPA/600/4-86/008. Las Vegas, NV: U.S. Environmental Protection
Agency, Office of Environmental Monioring and Quality Assurance.
Duncan, RA; Bethune, MG; Thayalakumaran, T; Christen, EW; McMahon, TA. (2008) Management of salt
mobilization in the irrigated landscape—A review of selected irrigation regions. J Hydrol 35 l(l-2):238-252.
Dunlop, J; McGregor, G; Horrigan, N. (2005) Potential impacts of salinity and turbidity in riverine ecosystems.
Characterization of impacts and a discussion of regional target setting for riverine ecosystems in Queensland. The
National Action Plan for Salinity and Water Quality Water Quality State-level Investment Project. Available online
at https://www.ehp.ald.gov.au/water/pdf/potential-impacts-sal-tur.pdf.
Dunlop, JE; Mann; RM; Hobbs, D; Smith, REW; Nanjappa, V; Vardy; Vink, S. (2015). Assessing the toxicity of
saline waters: the importance of accommodating surface water ionic composition at the river basin scale.
Australasian Bull Ecotoxicol Environ Chem 2: 1-15. Available online at
http://australasia.setac.org/wp-content/uploads/ABEEC-v2-pl.pdf.
Echols, BS; Currie, RJ; Cherry, DS. (2009) Influence of SC dissipation on benthic macroinvertebrates in the North
Fork Holston River, Virginia, downstream of a point-source brine discharge during severe low-flow conditions.
Hum Ecol Risk Assess 15(1): 170-184.
Echols, BS; Currie, RJ; Cherry, DS. (2010) Preliminary results of laboratory toxicity tests with the mayfly,
Isonychia bicolor (Ephemeroptera: Isonychiidae) for development as a standard test organism for evaluating streams
in the Appalachian coalfields of Virginia and West Virginia. Environ Monitor Assess 169(l-4):487-500.
Efron, B; Tibshirani, R. (1993) An introduction to bootstrap. Monographs on statistics and applied probability 57.
Boca Raton, FL: Chapman & Hall/CRC Press.
El-Ashry, MT; van Schilfgaarde, J; Schiffman, S. (1985) Salinity pollution from irrigated agriculture. J Soil Water
Conserv 40:48-52.
Entrekin, S; Evans-White, M; Johnson, B; Hagenbuch, E. (2011) Rapid expansion of natural gas development poses
a threat to surface waters. Front Ecol Environ 9:503-511.
Environment Canada and Health Canada. (2001) Priority substances list assessment report: road salts. Canadian
Environmental Protection Act, 1999. Available online at
http://wwl.prweb.com/prfiles/2008/02/07/370423/EnvironmentCanadareport.pdf.
Environment Canada. (2012) Federal contaminated sites action plan (FCSAP) ecological risk assessment guidance.
Prepared for U. Pott, Pacific and Yukon Environmental Stewardship Branch by Azimuth Consulting Vancouver,
BC, Canada. Available at http://geoenvirologic.ca/Documents/20120631 ERAGuidance Final En.pdf.
Epanchin, PN; Knapp, RA; Lawler, SP. (2010) Non-native trout impact an alpine nesting bird by altering
aquatic-insect subsidies. Ecology 91:2406-2415.
Erickson, RJ; Mount, DR; Highland, TL; Hockett, JR; Hoff, DJ; Jenson, CT; Norberg-King, TJ, Peterson, KN.
(2016) The acute toxicity of major ion salts to Ceriodaphnia dubia. II. Empirical relationships in binary salt
mixtures. Environ Toxicol Chem. 2016 Nov 1. doi: 10.1002/etc.3669. [Epub ahead of print].
8-4

-------
Evans, DH. (2008a) Osmotic and ionic regulation: cells and animals. Boca Raton, FL: CRC Press. 606 pp.
Evans, DH. (2008b) Teleost fish osmoregulation: what have we learned since August Krogh, Homer Smith, and
Ancel Keys? Am J Physiol Regul Intgr Comp Physiol 295(2):R704-R713.
Evans, DM; Zipper, CE; Donovan, PF; Daniels, WL. (2014) Long-term trends of specific conductance in waters
discharged by coal-mine valley fills in central Appalachia, USA. J Am Water Res Assoc 50(6): 1449-1460.
Evans, M; Frick, C. (2001) The effects of road salts on aquatic ecosystems. NWRI Contribution Series No. 02-308.
Saskatoon, Saskatchewan: National Water Research Institute (NWRI).
Evans-White, MA; Stelzer, RS; Lambert, GA. (2005) Taxonomic and regional patterns of benthic macroinvertebrate
elementals composition in streams. Freshw Biol 50:1786-1799.
Farag, AM; Harper, DD. (2012) The chronic toxicity of sodium bicarbonate defined under laboratory conditions. In:
Farag, AM; Harper, DD; eds. The potential effects of sodium bicarbonate, a major constituent of produced waters
from coalbed natural gas production, on aquatic life. Scientific Investigations Report 2012-5008; pp. 40-60.
Reston, VA: U.S. Geological Survey.
Feminella, JW. (1996) Comparison of benthic macroinvertebrate assemblages in small streams along a gradient of
flow permanence. J N Am Benthol Soc 15:651-669.
Findlay, SEG; Kelly, VR. (2011) Emerging indirect and long-term road salt effects on ecosystems. Ann NY Acad
Sci 1223:58-68.
Fisher, SG; Gray, LJ; Grimm, NB; Busch, DE. (1982) Temporal succession in a desert stream ecosystem following
flash flooding. Ecol Monogr 52:93-110.
Flotemersch, JE; Stribling, JB; Paul, MJ. (2006). Concepts and Approaches for the Bioassessment of Non-wadeable
Streams and Rivers. EPA 600-R-06-127. US Environmental Protection Agency, Cincinnati, Ohio.
https ://www3. epa. gov/re gionl /npdes/merrimackstation/pdfs/ar/AR-1157 .pdf.
Forman, RTT; Alexander, LE. (1998) Roads and their major ecological effects. Ann Rev Ecol Syst 29:207-31.
Fox, GA. (1991) Practical causal inference for ecoepidemiologist. J Toxicol Environ Health 33(4):359-374.
Fritz, KM; Fulton, S; Johnson, BR; Barton, CD; Jack, JD; Word, DA; Burke, RA. (2010) Structural and functional
characteristics of natural and constructed channels draining a reclaimed mountaintop removal and valley fill coal
mine. J N Am Benthol Soc 29:673-689.
Futuyma, DJ; Moreno, G. (1988) The evolution of ecological specialization. Ann Rev Ecol Syst 19:207-233.
Gaino, E; Bongiovanni, E. (1992) Comparative morphology of epithemata polar chorionic structures in the eggs of
Ephemerella ignita Ephemeroptera Ephemerellidae. Trans Am Microscop Soc 111(3):255-265.
Geckler, JR; Horning, WB; Neiheisel, TM; Pickering, QH; Robinson, EL. (1976) Validity of laboratory tests for
predicting copper toxicity in streams. EPA-600/3-76-116. Deluth MN: Environmental Research Laboratory.
Available online at http://oaspub.epa.gov/eims/eimscomm.getfile7p download id=496962.
Gerritsen, J; Burton, J; Barbour, MT. (2000) A stream condition index for West Virginia wadeable streams.
Prepared by: Tetra Tech, Inc. Owing Mills, MD. Prepared for: U.S. EPA Region 3 Environmental Services
Division, and U.S. EPA Office of Science and Technology, Office of Water. Available online at
http://www.dep.wv.gov/WWE/watershed/bio fish/Documents/WVSCI.pdf.
Gerritsen, J; Zheng, L; Burton, J; Boschen, C; Wilkes, S; Ludwig, J; Cormier, S. (2010) Inferring causes of
biological impairment in the Clear Fork Watershed, West Virginia. EPA/600/R-08/146. Washington, DC: U.S.
Environmental Protection Agency, Office of Research and Development, National Center for Environmental
Assessment. Available online at http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=201963.
8-5

-------
Gesser, H; Poupa, O. (1983) Acidosis and cardiac muscle contractility: comparative aspects. Comp Biochem Physiol
A Comp Physiol 76(3):559-66.
Gibbs, RJ. (1970) Mechanisms controlling world water chemistry. Science 170:1088-1090.
Gillis, PL. (2011) Assessing the toxicity of sodium chloride to the glochidia of freshwater mussels: Implications for
salinization of surface waters. Environ Pollut 159(6): 1702-1708.
Goddard, MA; Mikhailova, EA; Post, CJ; Schlautman, MA. (2007) Atmospheric Mg2+ wet deposition within the
continental United States and implications for soil inorganic carbon sequestration. Tellus. 59B:50-56.
Godwin KA; Hafner, SD; Buff, MF. (2003) Long-term trends in sodium and chloride in the Mohawk River, New
York: The effect of fifty years of road-salt application. Environ Pollut 124(2):273-281.
Goldstein, JN; Woodward, DF; Farag, AM. (1999) Movement of adult Chinook salmon during spawning migration
in a metals-contaminated system, Cour d'Alene River, Idaho. Trans Am Fish Soc 128:121-129.
Gonzalez, RJ; McDonald, DG. (1992) The relationship between oxygen consumption and ion loss in a freshwater
fish. J Exp Biol 163(1):317-32.
Gray, LJ. (1993) Emergence production and export of aquatic insects in riparian habitats in a tallgrass prairie. Am
Midi Nat 129:288-300.
Gregory, KB; Vidic, RD; Dzombak, DA. (2011) Water management challenges associated with the production of
shale gas by hydraulic fracturing. Elements 7(3): 181—186.
Griffith, MB. (2014) Natural variation and current reference for specific SC and major ions in wadeable streams of
the coterminous U.S. Freshw Sci 33(1): 1-17.
Griffith, MB. 2016. Toxicological perspective on the osmoregulation and ionoregulation physiology of major ions
by freshwater animals: Teleost fish, crustacea, aquatic insects, and Mollusca: Osmoregulation and ionoregulation
physiology of major ions. Environmental Toxicology and Chemistry. DOI: 10.1002/etc.3676.
Griffith, MB; Norton, SB; Alexander, LC; Polland, AL; LeDuc, SD. (2012) The effects of mountaintop mines and
valley fills on the physicochemical quality of stream ecosystems in the central Appalachians: a review. Sci Total
Environ 417-418:1-12.
Grubbs, SA. (2010) Influence of flow permanence on headwater macroinvertebrate communities in a Cumberland
Plateau watershed, USA. Aqua Ecol 45:185-195.
Hall, RO, Jr; Peredney, CL; Meyer, JL. (1996) The effect of invertebrate consumption on bacterial transport in a
mountain stream. Limnol Oceanogr 41:1180-1187.
Haluszczak, LO; Rose, AW; Kump, LR (2013) Geochemical evaluation of flowback brine from Marcellus gas wells
in Pennsylvania, USA. Appl Geochem 28:55-61.
Hamburg, SP; Yanai, RD; Arthur, MA; Blum, JD; Siccama, TG. (2003) Biotic control of calcium cycling in
northern hardwood forests: Acid rain and aging forests. Ecosystems 6(4):399-406.
Harper, DD; Farag, AM; Skaar, D; Selch, T. (2012) The acute toxicity of sodium bicarbonate defined under
laboratory conditions. In: Farag, AM; Harper, DD; eds. The potential effects of sodium bicarbonate, a major
constituent of produced waters from coalbed natural gas production, on aquatic life. Scientific Investigations Report
2012-5008. U.S. Geological Survey, Reston, VA; pp. 7-39.
Hastie, T; Tibshirani, R. (1986) Generalized additive models. Stat Sci 1(3):297—318.
Hawkins, CP; Cao, Y; Roper, B. (2010) Method of predicting reference condition biota affects the performance and
interpretation of ecological indices. Freshw Biol 55(5): 1066-85.
8-6

-------
Heckman, CW. (1983) The recovery of the biotic community in a lotic freshwater habitat after extensive destruction
by chlorine. IntRev Gesamten Hydrobiol 68:207-226.
Hem, JD. (1985) Study and interpretation of the chemical characteristics of natural water. U.S. Geological Survey
Water-Supply Paper 2254. Richmond, VA: USGS. Available online at
http://pubs.usgs.gov/wsp/wsp2254/html/pdf.html.
Herlihy, AT; Paulsen, SG; Sickle, JV; Stoddard, JL; Hawkins, CP; Yuan, LL. (2008) Striving for consistency in a
national assessment: the challenges of applying a reference-condition approach at a continental scale. J N Am
Benthol Soc 27(4):860-877.
Higgins, CL; Wilde, GR. (2005) The role of salinity in structuring fish assemblages in a prairie stream system.
Hydrobiologia 549:197-203.
Hill, AB. (1965) The environment and disease: Association or causation? Proc Royal Soc Med 58(5):295-300.
Hille, B. (2001) Ion channels of excitable membranes, 3rd edition. Sunderland, MA: Sinauer Associates, Inc.
Hitt, NP; Chambers, DB. (2014) Temporal changes in taxonomic and functional diversity of fish assemblages
downstream from mountaintop mining. Freshw Sci 33(3):915-926. Available online at
http://www.istor.org/stable/10.1086/676997.
Hooper, DU; Chapin, FS, III; Ewel, JJ; Hector, A; Inchausti, P; Lavorell, S; Lawton, JH; Lodge, DM; Loreau, M;
Naeem, S; Schmid, B; Setalia, H; Symstad, AJ; Vandermeer, J; Wardle, DA. (2005) Effect of biodiversity on
ecosystem functioning: A consensus of current knowledge. Ecol Monogr 75:3-35.
Hopkins, RL, II; Roush, JC. (2013) Effects of mountaintop mining on fish distributions in central Appalachia. Ecol
Freshw Fish 22(4):578-586.
Huntington, TG; Hooper, RP; Aulenbach, BT. (1994) Hydrologic processes controlling sulfate mobility in a small
forested watershed. Water Res Res 30(2):283-295.
Huryn, AD; Wallace, JB. (2000) Life history and production of stream insects. Ann Rev Entomol 45:83-110.
Hynes, HBN. (1960) The biology of polluted waters. Liverpool, UK: Liverpool University Press.
Ingebritsen, SE; Gelwick, KD; Randolph-Flagg, NG; Crankshaw, IM; Lundstrom, EA; McCulloch, CL; Murveit,
AM; Newman, AC; Mariner, RH; Bergfeld, D; Tucker, DS; Schmidt, ME; Spicer, KR; Mosbrucker, AR; Evans,
WC. (2014) Hydrothermal monitoring data from the Cascade Range, northwestern United States: U.S. Geological
Survey Data Set. Washington, DC: U.S. Department of the Interior, U.S. Geological Survey,
doi: 10.5066/F72N5088.
Jaffe, LF. (1991) The path of calcium in cytosolic calcium oscillations: A unifying hypothesis. Proc Natl Acad Sci
USA. 88:9883-9887.
Jackson, RB; Jobbagy, EG. (2005) From icy roads to salty streams. Natl Acad Sci USA 102:14487-14488.
Jackson, RE; Reddy, KJ. (2007) Geochemistry of coalbed natural gas (CBNG) produced water in Powder River
Basin, Wyoming: salinity and sodicity. Water Air Soil Poll 184:49-61.
Johnson, BR; Haas, A; Fritz, KM. (2010) Use of spatially explicit physiochemical data to measure downstream
impacts of headwater streams. Water Res Res 46:W09526.
Johnson, BL; Johnson, MK. (2015) An evaluation of a field-based aquatic life benchmark for specific conductance
in northeast Minnesota. Prepared for Water Legacy. Available at
http://waterlegacv.org/sites/default/files/u42412/Ex.16 JohnsonMNConductivitvEvaluationRpt%26Attachments(No
v.%202015').pdf.
8-7

-------
Johnson, BR; Weaver, PC; Nietch, CT; Lazorchak, JM; Straewing, KA; Funk, DH. (2015) Elevated major ion
concentrations inhibit larval mayfly growth and development. Environ Toxicol Chem 34(1): 167-72.
Johnson, RC; Jin, H-S; Carreiro, MM; Jack, JD. (2013) Macroinvertebrate community structure, secondary
production and trophic-level dynamics in urban streams affected by non-point-source pollution. Freshw Biol
58:843-857.
Karatayev, AY; Miller, TD; Burlakova, LE. (2012) Long-term changes inunionid assemblages in the Rio Grande,
one of the World's top 10 rivers at risk. Aqua Conserv: Mar Freshw Ecosyst 22:206-219.
Kattwinkel, M; Rombke, J; Liess, M. (2012) Ecological recovery of populations of vulnerable species driving the
risk assessment of pesticides. Support Publ 2012: EN-338. 98 pp. Available online at:
http://www.efsa.europa.eu/en/efsaiournal/doc/338e.pdf.
Kaushal, SS; Groffman, PM; Likens, GE; Belt, KT; Stack, WP; Kelly, VR; Band, LE; Fisher, GT. (2005) Increased
salinization of fresh water in the northeastern United States. Proc Natl Acad Sci USA 102(38): 13517-13520.
Kaushal, SS; Likens, GE; Utz, RM; Pace, ML; Grese, M; Yespen, M. (2013) Increased river alkalinization in the
Eastern U.S. Environ Sci Technol 17(18): 10302—10311.
Kelly, VR; Lovett, GM; Weathers, KC; Findlay, SEG; Strayer, DL; Burns, DI; Likens, GE. (2008) Long-term
sodium chloride retention in a rural watershed: legacy effects of road salt on streamwater concentration. Environ Sci
Technol 42(2):410-415.
Kelting, DL; Laxson, CL; Yerger, EC. (2012) Regional analysis of paved roads on sodium and chloride in lakes.
Water Res 46:2749-2758.
Kennedy, AJ; Cherry, DS; Currie, RJ. (2003) Field and laboratory assessment of a coal processing effluent in the
Leading Creek watershed, Meigs County, Ohio. Arch Environ Contam Toxicol 44(3):324-331.
Kennedy, AJ; Cherry, DS; Currie, RJ. (2004) Evaluation of ecologically relevant bioassays for a lotic system
impacted by a coal-mine effluent, using Isonychia. Environ Monit Assess 95(l-3):37-55.
Kennedy, AJ; Cherry, DS; Zipper, CE. (2005) Evaluation of ionic contribution to the toxicity of a coal-mine effluent
using Ceriodaphnia dubia. Arch Environ Contam Toxicol 49(2): 155-162.
Kimmel, WG; Argent, DG. (2010) Stream fish community responses to a gradient of SC. Water Air Soil Pollut
206(l-4):49-56.
Komnick, H. (1977) Chloride cells and chloride epithelia of aquatic insects. Int Rev Cytol 49:285-328.
Krug, EC; Frink, CR. (1983) Acid rain on acid soil: A new perspective. Science. 221:520-525.
Kunz, JL; Conley, JM; Buchwalter, DB; Norber-King, TJ; Kemble, NE; Wang, N; Ingersoll, CG. (2013) Use of
reconstituted waters to evaluate effects of elevated major ions associated with mountaintop coal mining on
freshwater invertebrates. Environ Toxicol Chem 32(12):2826-2835.
Larson, AG; Marti, PB. (1996) Relationship between ground water and surface water in the Quilceda Creek
watershed. Environmental Information. Publication No. 96-333. Pp. 33. Olympia, WA: Washington State
Department of Ecology. Available online at https://fortress.wa. gov/ecv/publications/documents/963 3 3 .pdf.
Lasier, PJ; Hardin, IR. (2010) Observed and prediction reproduction of Ceriodaphnia dubia exposed to chloride,
sulfate, and bicarbonate. Environ Toxicol Chem 29(2):347-358.
Lefebvre, O; Moletta, R. (2006) Treatment of organic pollution in industrial saline wastewater: A literature review.
Water Res 40(20):3671-3682.

-------
Legendre, P. (1998) Model II regression user's guide, R edition. R Vignette.
https://www.google.com/7gws rd=ssl#q=model+II+regression.
Leland, HV; Brown, LR; Mueller, DK. (2001) Distribution of algae in the San Joaquin River, California, in relation
to nutrient supply, salinity and other environmental factors. Freshw Biol 46:1139-1167.
Lenat, DR. (1984) Agriculture and stream water quality: A biological evaluation of erosion control practices.
Environ Manage 8(4):333-343.
Lerotholi, S; Palmer, CG; Rowntree, K. (2004) Bioassessment of a river in a semi-arid, agricultural catchment,
Eastern Cape. In: Proceedings of the 2004 Water Institute of Southern Africa (WISA) Biennial Conference, Cape
Town, South Africa, pp. 338-344.
Liess, M; Beketov, M. (2011) Traits and stress: Key to identify community effects of low levels of toxicants in test
systems. Ecotoxicology. 20(6): 1328-1340.
Liess, M; Schulz, R. (1999) Linking insecticide contamination and population response in an agricultural stream.
Environ Toxicol Chem 18(9): 1948-1955.
Likens, GE; Driscoll, CT; Buso, DC; Siccama, TG; Johnson, CE; Lovett, GM; Fahey, TJ; Reiners, WA; Ryan, DF;
Martin, CW; Bailey, SW. (1998) The biogeochemistry of calcium at Hubbard Brook. Biogeochemistry. 41:89-173.
Likens, GE; Bormann, FH; Johnson, NM; Fisher, DW; Pierce, RS. (1970) Effects of forest cutting and herbicide
treatment on nutrient budgets in the Hubbard Brook watershed-ecosystem. Ecol Monogr 40(l):23-47.
Lindberg, TT; Bernhardt, ES; Bier, R; Helton, AM; Merola, RB; Vengosh, A; Di Giulio, LT. (2011) Cumulative
impacts of mountaintop mining on an Appalachian watershed. Proc Natl Acad Sci 108(52):20929-20934.
Lynch, JA; Corbett, ES. (1989) Hydrologic control of sulfate mobility in a forested watershed. Water Res Res
25(7):1695-1703.
MacArthur, R; Levins, R. (1967) The limiting similarity, convergence, and divergence of coexisting species. Am
Natural 101(921):377-385.
Marshall, WS. (2002) Na+, CI . Ca2+ and Zn2+ transport by fish gills: retrospective review and prospective synthesis.
JExpZool 293:264-283.
Mayhugh, J; Ziemkiewicz, P. (2005) Treatment of chloride contaminated mine water in NW West Virginia. USGS
West Virginia Water Resources Investigations Project Report WRI 54. Available from:
http://water.usgs.gov/wrri/04grants/Progress%20Completion%20Reports/2003WV16B.pdf.
McCleskey, RB. (2011) Electrical conductivity of electrolytes found in natural waters from (5 to 90 °C). J Chem
Engineer Data 56:317-327.
McNab, WH; Avers, PE. (1994) Ecological subregions of the United States. Washington, DC: USDA Forest
Service.
Mebane, CA; Eakins, RJ; Fraser, BG; Adams, WJ. (2015) Recovery of a mining-damaged stream ecosystem.
Elementa: Sci Anthrop. Pp. 34 Available on line at: https://www.elementascience.org/articles/42.
Merriam, ER; Petty, JT; Merovich, GT; Fulton, JB; Strager, MP. (2011) Additive effects of mining and residential
development on stream conditions in a central Appalachian watershed. J N Am Benthol Soc 30(2):399-418.
Merricks, TC; Cherry, DS; Zipper, CE; Currie, RJ; Valenti, TW. (2007) Coal-mine hollow fill and settling pond
influence on headwater streams in southern West Virginia, USA. Environ Monit Assess 129(l-3):359-378.
Merritt, RW; Cummins, KW. (1996) An introduction to the aquatic insects of North America. 3rd Edition. Dubuque,
IA: Kendall/Hunt. 862 pp.
8-9

-------
Meyer, JS; Sanchez, DA; Brookman, JA; McWorter, DB; Bergman, HL. (1985) Chemistry and aquatic toxicity of
raw oil shale leachates from Piceance Basin, Colorado. Environ Toxicol Chem 4(4):559-572.
Micieli, MV; Risso, P; Achinelly, MF; Villar, MM; Muttis, E. (2012) Population dynamics between the mermithid
Strelkovimermis spiculatus (Nematoda: Mermithidae) and the floodwater mosquito Ochlerotatus albifasciatus
(Diptera: Culicidae) over time. Biol Contr 61:55-63.
Minshall, GW; Andrews, DA; Manuel-Faler, CY. (1983) Application of island biogeographic theory to streams:
Macroinvertebrate recolonization of the Teton River, Idaho. In: Barnes, JR; Minshall, GW; eds. Stream ecology:
application and testing of general ecological theory. Pp 279-297. New York: Plenum Press.
Molles, MC. (1985) Recovery of a stream invertebrate community from a flash flood in Tesque Creek, New Mexico.
Southwest Nat 30:279-287.
Mount, DR; Erickson, RJ; Highland, TL; Hockett, JR; Hoff, DJ; Jenson, CT; Norberg-King, TJ; Peterson, KN; y M.
Polaske,ZM; Wisniewskiz, S. (2016) The acute toxicity of major ion salts to Ceriodaphnia dubia: I. influence of
background water chemistry. Environ Toxicol Chem 35(12):3039-3057.
Mount, DR; Gulley, DD; Hockett, JR; Garrison, TD; Evans, JM. (1997) Statistical models to predict the toxicity of
major ions to Ceriodaphnia dubia, Daphnia magna, and Pimephalespromelas (fathead minnows). Environ Toxicol
Chem 16(10):2009-2019.
Mueller, DK; Martin, JD; Lopes, TJ. (1997) Quality-control design for surface-water sampling in the National
Water-Quality Assessment Program. (Open-file report 97-223). Denver, CO: US Geological Survey.
https://pubs.usgs.gov/of/1997/223/pdf/ofr97-223.pdf.
Mullins, GL; Hansen, DJ. (2006) Chapter 4: Basic soil fertility. In: Evanylo, G; ed. Mid-Atlantic nutrient
management handbook. Mid-Atlantic Regional Water Program.
Mulrennan, ME; Woodroffe, CD. (1998) Saltwater intrusion into the coastal plains of the Lower May River,
Northern Territory, Australia. J Environ Manage 54(3): 169-188.
NADP (National Atmospheric Deposition Program) (2013) NADP maps and data. Available online at:
http://nadp.sws.uiuc.edu/data/.
Nakano, S; Murakami, M. (2001) Reciprocal subsidies: Dynamic interdependence between terrestrial and aquatic
food webs. Proc Natl Acad Sci 98:166-70.
NRC (National Research Council). (2010) Management and effects of coalbed methane produced water in the
western United States. Committee on Management and Effects of Coalbed Methane Development and Produced
Water in the Western United States. Washington, D.C.: National Academies Press.
Nelson, D; Cox, M. (2005) Lehninger principles of biochemistry. 4th edition. New York: W. H. Freeman & Co.; pp.
395-397.
Newbold, JD; Elwood, JW; O'Neill, RV; Sheldon, AL. (1983) Phosphorus dynamics in a woodland stream
ecosystem: A study of nutrient spiraling. Ecology 64:1249-1263.
Newbold, JD; O'Neill, RV; Elwood, JW; van Winkle, W. (1982) Nutrient spiraling in streams: Implications for
nutrient limitation and invertebrate activity. Am Natural 120:628-652.
Newman, MC; Ownby, DR; Mezin, LCA; Powell, DC; Christensen, TRL; Lerberg, SB; Anderson, B-A. (2000)
Applying species sensitivity distributions in ecological risk assessment: Assumptions of distribution type and
sufficient number of species. Environ Toxicol Chem 19(2):508-515.
8-10

-------
Newman, MC; Ownby, DR; Mezin, LCA; Powell, DC; Christensen, TRL; Lerberg, SB; Anderson, B-A; Padma,
TV. (2001) Species sensitivity distributions in ecological risk assessment: Distributional assumptions, alternative
bootstrap techniques, and estimation of adequate number of species. In: Posthuma, L; Suter, GW, II; Traas, TP; eds.
Species sensitivity distributions in ecotoxicology. pp 119-132. Boca Raton, FL: CRC Press, LLC.
Nezat, CA; Blum, JD; Yanai, RD; Park, BB. (2008) Mineral sources of calcium and phosphorus in soils of the
Northeastern United States. Soil Sci Soc Am J 72:1786-1794.
Niemi, GJ; DeVore, P; Detenbeck, Taylor, D; Yount, JD; Lima, A; Pastor, J; Naiman, RJ. (1990) Overview of case
studies on recovery of aquatic ecosystems from disturbance. Environ Manage 14:571-588.
Nietch, C; Lazorchak, J; Johnson, B; Allen, J. (2014) Evaluation of total dissolved solids and specific conductance
water quality targets with paired single-species and mesocosm community exposures. Presented at Society of
Environmental Toxicology and Chemistry North America 35th Annual Meeting, Vancouver, BC, CANADA,
November 09-13, 2014. Available online at
http://cfpub.epa.gov/si/si public record Report.cfm?dirEntrvId=293057&CFID=17865572&CFTOKEN=51190896
Norton, SB; Cormier, SM; Suter, GW II. (2015) Ecological Causal Assessment. CRC Press, Taylor and Francis
Group, Boca Raton, FL.
Novotny, EV; Murphy, D; Stefan, HG. (2008) Increase of urban lake salinity by road deicing salt. Sci Total Environ
406:131-144.
NRC (National Research Council). (2010) Management and effects of coalbed methane produced water in the
western United States. Committee on Management and Effects of Coalbed Methane Development and Produced
Water in the Western United States. Washington, D.C.: National Academies Press.
NYSDEC (New York State Department of Environment and Conservation). (2000). Memorandum from Scott
Kisbaugh to Jay Bloomfield regarding reference lakes for nutrient criteria. September 26, 2000.
Olson, JR. (2012) The influence of geology and other environmental factors on stream water chemistry and benthic
invertebrate assemblages. PhD Dissertation, Utah State University, Logan, Utah. Available at
http://digitalcommons.usu.edu/cgi/viewcontent.cgi?article=2340&context=etd.
Olson, JR; Hawkins, CP. (2012) Predicting natural base-flow stream water chemistry in the western United States.
Water Resour Res 48:W02504.
Omernik, JM. (1987) Ecoregions of the conterminous United States. Ann Assoc Am Geograph 77:118-125.
Omernik, JM. (1995) Ecoregions: a spatial framework for environmental management. In: Davis, WS; Simon, TP;
(eds). Biological assessment and criteria: tools for water resource planning and decision making. [Pp 49-62], Boca
Raton, FL: Lewis Publishers.
Oregon DEQ (Department of Environmental Quality). (2009) Water monitoring and assessment mode of operations
manual (MOMs). [DEQ03-LAB-0036-SOP]. Hillsboro, Oregon: Laboratory and Environmental Assessment
Devision. Available online at http://www.dea.state.or.us/lab/techrpts/docs/DE003LAB0036SQP.pdf.
OWEB (Oregon Watershed Enhancement Board). (2001) Draft Oregon watershed assessment manual. Appendix A:
Ecoregion descriptions. Salem, OR: OWEB. Available online at
http://www.oregon.gov/OWEB/docs/pubs/wa manual99/apdxl-ecoregions.pdf.
Palmer, MA; Bernhart, ES; Schlesinger, WH; Eshleman, KN; Fourfoula-Georgiou, E; Hendryx, MS; Lemly, AD;
Likens, GE; Loucks, OL; Power, ME; White, PS; Wilcock, PR. (2010) Mountaintop mining consequences. Science
327:148-149.
Paul, MJ; Meyer, JL. (2001) Streams in the urban landscape. Ann Rev Ecol Syst 32:333-365.
8-11

-------
Percival, E; Whitehead H. (1928) Observations on the ova and oviposition of certain Ephemeroptera and Plecoptera.
Proc Leeds Phil Soc (Sci Sect) 1:271-288.
Peterson, AT, Soberon, J; Pearson, RG; Anderson, RP; Martinex-Meyer, E; Nakamura, M; Araujo, MB. (2011)
Ecological niches and geographic distributions. Princeton, NJ: Princeton University Press.
Pond, GJ. (2004) Effects of surface mining and residential land use on headwater stream biotic integrity in the
eastern Kentucky coalfield region. Kentucky Department of Environmental Protection, Division of Water, Frankfort,
KY.
Pond, GJ. (2010) Patterns of Ephemeroptera taxa loss in Appalachian headwater streams (Kentucky, USA)
Hydrobiologia 641(1):185—201.
Pond, GJ; Passmore, ME; Borsuk, FA; Reyolds, L; Rose, CJ. (2008) Downstream effects of mountaintop coal
mining: comparing biological conditions using family- and genus-level macroinvertebrate bioassessment tools. J N
Am Benthol Soc 27(3):717-737.
Pond, GJ; Passmore, ME; Pointon, ND; Felbinger, JK; Walker, CA; Krock, KJ; Fulton, JB; Nash, WL. (2014)
Long-term impacts on macroinvertebrates downstream of reclaimed mountaintop mining valley fills in Central
Appalachia. Environ Manage 54(4):919-933. Available online at http://www.ncbi.nlm.nih.gov/pubmed/24990807.
Posthuma, L; Suter GW, II; Traas, TP; eds. (2001) Species sensitivity distributions in ecotoxicology. Boca Raton,
FL: Lewis Publishers.
Potapova, M. (2005) Relationships of soft-bodied algae to water-quality and habitat characteristics in the U.S. rivers:
Analysis of the National Water-Quality Assessment (NAWQA) Program data set. Philadelphia, PA: The Academy
of Natural Sciences, http://diatom.acnatsci.org/autecology/uploads/Report Qctober20 .pdf.
Potapova, M. (2014) Autecology of freshwater algae datafiles and links to internet resources. Philadelphia, PA: The
Academy of Natural Sciences of Drexel University, http ://diatom. acnatsci. org/autecolo gy/.
Potapova, M; Charles, DF. (2003) Distribution of benthic diatoms in U.S. rivers in relation to SC and ionic
composition. Freshw Biol 48:1311-1328.
Price, JE; Zipper, CE; Jones, JW; Franck, CT. (2014) Water and sediment quality in the Clinch River, Virginia and
Tennessee, USA, over nearly five decades. J Am Water Res Assoc 50(4):837-858. DOI: 10.1111/jawr. 12219.
R Development Core Team (2011) R: A language and environment for statistical computing. Vienna, Austria: R
Foundation for Statistical Computing. Available online at http://www.R-proiect.org.
Remane, A. (1971) Ecology of brackish water. In: Remane A; Schlieper, C; eds. Biology of brackish water. New
York, NY: John Wiley and Sons.
Resh, VH. (1995) Freshwater benthic macroinvertebrates and rapid assessment procedures for water quality
monitoring in developing and newly industrialized countries. In: Davis, WS; Simon, TP; eds. Biological assessment
and criteria, pp. 167-177. Boca Raton, FL: Lewis Publishers.
Richardson, JS. (1993) Limits of productivity of streams: Evidence from studies of macroinvertebrates. Can Spec
Pub Fish Aqua Sci 118:9-15.
Rose, S. (2007). The effects of urbanization on the hydrochemistry of base flow within the Chattahoochee River
Basin (Georgia, USA) J Hydrol 341:42-54.
Rubach, MN; Ashauer, R; Buchwalter, DB; DeLange, H; Hamer, M; Preuss, TG; Topke, K; Maund, SJ. (2011)
Framework for traits-based assessment in ecotoxicology. Integr Environ Assess Manag 7:172-186.
8-12

-------
Ruhl, L; Vengosh, A; Dwyer, GS; Hsu-Kim, H; Schwartz, G; Romanski, A; Smith, SD. (2012) The impact of coal
combustion residue effluent on water resources: A North Carolina example. Environ Sci Technol 46
(21):12226-12233.
Samarina, VP. (2007) Assessment of the impact of economic activity on the degree of overland flow contamination
in the zone of the Kursk-Belgorod magnetic anomaly: Case study of the Oskol River. Water Resour 34(5):549-553.
Sams, JI; Beer, KM. (2000) Effects of coal-mine drainage on stream water quality in the Allegheny and
Monongahela River basins - sulfate transport and trends. USGS Water Resources Investigations Report 99-4208.
Lemoyne, PA: National Water-Quality Assessment Program. Available online at
http://pa.water.usgs.gov/reports/wrir 99-4208.pdf.
Scanlon, BR; Stonestorm, DA; Reedy, R; Leaney, FW; Gates, JB; Cresswell, G. (2009) Inventories and mobilization
of unsaturated zone sulfate, fluoride, and chloride related to land use change in semiarid regions, southwestern
United States and Australia. Water Resour Res 45:W00A18.
Schlesinger, WH. (1997) Biogeochemistry, an analysis of global change, Second Edition. San Diego, CA: Academic
Press. 588 pp.
Schofield, K; Ziegler, R. (2010) Detailed conceptual model diagram for ionic strength, (modified 7/10). U.S.
Environmental Protection Agency, Causal Analysis/Diagnosis Decision Information System. Available online at:
http://www.epa.gov/caddis/pdf/conceptual model/ion CD det.pdf.
Sheehan, PJ; Winner, RW. (1984) Comparison of gradient studies in heavy-metal-polluted streams. In: Sheehan, PJ;
Miller, DR; Butler, GC; Bourdeau, PJ; eds. Effects of pollutants at the ecosystem level, pp. 255-271. Oxford,
England: John Wiley and Sons, Ltd.
Smith, DG. (2001) Pennak's freshwater Invertebrates of the United States: Porifera to Crustacea, 4th edition. New
York, NY: John Wiley & Sons Inc. 648 pp.
Smith, JM; Dornish, M; Wood, EJ. (2005) Involvement of protein kinase C in chitosan glutamate-mediated tight
junction disruption. Biomaterials. 26(16):3269-76.
Smith, RF; Alexander, LC; Lamp, WO. (2009) Dispersal by terrestrial stages of stream insects in urban watersheds:
a synthesis of current knowledge. J N Am Benthol Soc 28(4): 1022-1037.
Smithson, J. (2007) West Virginia stream/river survey design 2007—2 111. Charleston, WV. West Virginia
Department of Environmental Protection, Division of Water and Waste Management.
Soucek, DJ; Kennedy, AJ. (2005) Effects of hardness, chloride, and acclimation on the acute toxicity of sulfate to
freshwater invertebrates. Environ Toxicol Chem 24(5): 1204-1210.
Soucek, DJ; Dickinson, A. (2015) Full-life chronic toxicity of sodium salts to the mayfly Neocloeon triangulifer in
tests with laboratory cultured food. Environ Toxicol Chem 34(9):2126-37.
Stallard, RF; Edmond, JM. (1987) Geochemistry of the Amazon. 3. Weathering chemistry and limits to dissolving
inputs. J Geophys Res 92:8293-8302.
Stauffer, JR; Ferreri, CP. (2002) Characterization of stream fish assemblages in selected regions of mountain top
removal/valley fill coal mining. In: Draft programmatic environmental impact statement on mountaintop
mining/valley fills in Appalachia (2003) Appendix D. U.S. Environmental Protection Agency, Region 3,
Philadelphia, PA.
Steffy, LU; McGinty, AL; Welty, C; Kilham, SS. (2004) Connecting ground water influxes with fish species
diversity in an urbanized watershed. J Am Water Resour Assoc 40(5): 1269-1275.
Stevens, DL; Olsen, AR. (2004) Spatially balanced sampling of natural resources. J Am Stat Assoc 99:262-278.
8-13

-------
Stewart, MA; Bhatt, TN; Fennema, RJ; Fitterman, DV. (2002) The road to Flamingo: an evolution of flow pattern
alterations and salinity instrusion into the Lower Glades, Everglades National Park. USGS Open-File Report 02-59.
Denver, CO: U.S. Geological Survey. Available online at
http://www.researchgate.net/publication/237232996 The Road to Flamingo an Evaluation of Flow Pattern Alt
erations and Salinity Intrusion in the Lower Glades Everglades National Park.
Stoddard, JL; Larsen, DP; Hawkins, CP; Johnson, RK; Norris, RH. (2006) Setting expectations for the ecological
condition of streams: The concept of reference condition. Ecol Appl 16(4): 1267-1276.
Stout, B; Wallace, JB. (2003) A survey of eight major aquatic insect orders associated with small headwater streams
subject to valley fills from mountaintop mining. Appendix in Mountaintop Mining/Valley Fills in Appalachia. Final
Programmatic Environmental Impact Statement. Philadelphia, PA: U.S. Environmental Protection Agency.
Sutcliffe, DW. (1962) The composition of haemolymph in aquatic insects. Exp Biol 39:325-343.
Sutcliffe, DW. (1974) Sodium regulation and adaptation to fresh-water in isopod genus Asellus. J Exp Biol
61(3):719—736.
Suter, GW, II. (2007) Ecological risk assessment, 2nd edition. Boca Raton, FL: CRC Press.
Suter, GW, II; Cormier, SM. (2008) What is meant by risk-based environmental quality criteria? Integr Environ
Assess Manag 4(4):486-489.
Suter, GW, II; Cormier, SM. (2013) A method for assessing the potential for confounding applied to ionic strength
in Central Appalachian streams. Environ Toxicol Chem 32(2):288-295.
Suter, GW, II; Cormier, S.M. (2015) Why care about aquatic insects: uses, benefits and services. Integr Environ
Assess Manage 11(2): 188—194. http://onlinelibrarv.wilev.com/doi/10.1002/ieam. 16/pdf.
Suter, GW, II; Traas, T; Posthuma, L. (2001) Issues and practices in the derivation and use of species sensitivity
distributions. In: Posthuma, L; Suter, GW, II; Traas, T; eds. Species sensitivity distributions in ecotoxicology. Boca
Raton, FL: Lewis Publishers; pp. 437-474.
Sweka, JA; Hartman, KJ. (2008) Contribution of terrestrial invertebrates to yearly brook trout consumption and
growth. Trans Am Fish Soc 137:224-235.
Svendsen, CR; Quinn, T; Kolbe, D. (2004) Review of macroinvertebrate drift in lotic ecosystems. Final report for
Wildlife Research Program Environmental and Safety Division. Olympia, WA: Washington Department of Fish and
Wildlife.
TDEC (Tennessee Department of Environment and Conservation) (2000) Letter to Geoff Grubbs containing
comments on draft nutrient criteria recommendation. October 5, 2000.
Thomas, LP. (2002) Coal geology, 1st ed. West Sussex, England: John Wiley and Sons, Inc. 379 pp.
Thorp, JH; Covich, AP. (2001) Ecology and classification of North American freshwater invertebrates. New York,
NY: Academic Press. 1056 pp.
Timpano, A; Schoenholtz, S; Zipper, C; Soucek, D. (2011) Levels of dissolved solids associated with aquatic life
effects in headwater streams of Virginia's Central Appalachian coalfield region. Report prepared for the Virginia
Department of Environmental Quality; Virginia Department of Mines, Minerals, and Energy; and the Powell River
Project. Prepared by Virginia Tech. Blacksburg, VA. Available online at:
http://www.prp.cses.vt.edu/Research Results/Timpano TDSReport 2011.pdf.
Timpano, AJ; Schoenholtz, SH; Soucek, DJ; Zipper, CE. (2014) Salinity as a limiting factor for biological condition
in mining influenced Central Appalachian headwater streams. J Am Water Resour Assoc 5 l(l):240-250.
8-14

-------
U.S. EPA (Environmental Protection Agency). (1982) Conductance (specific conductance, (imhos at 25°C) Method
#: 120.1. Washington, DC: Office of Science and Technology. Available online at
https://www.epa.gov/sites/production/files/2015-08/documents/method 120-1 1982.pdf.
U.S. EPA (Environmental Protection Agency). (1983) Water quality standards handbook. Washington, DC: Office
of Water Regulations and Standards. EPA/823/R083/002. Available online at
https ://nepis. epa. gov/Exe/ZvPDF.c gi?Dockev=2000UIP S .PDF
U.S. EPA (Environmental Protection Agency). (1985) Guidelines for deriving numeric National Water Quality
Criteria for the protection of aquatic organisms and their uses. Washington, DC: Office of Research and
Development, Environmental Research Laboratories. PB85-227049. Available online at
http://water.epa.gov/scitech/swguidance/standards/upload/2009 01 13 criteria 85guidelines.pdf.
U.S. EPA (Environmental Protection Agency). (1987) Handbook of methods for acid deposition studies: Laboratory
analyses for surface water chemistry. EPA/600/4-87/026. Washington, DC: U.S. Environmental Protection Agency,
Office of Research and Development, Acid Deposition and Atmospheric Research Division. Available online at
http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=30000TA0.txt.
U.S. EPA (Environmental Protection Agency). (1991) Technical support document for water quality-based toxics
control. Washington, DC: Office of Science and Technology, Office of Water. EPA/505/2-90/001. Available
online at https://www3.epa.gov/npdes/pubs/owm0264.pdf.
U.S. EPA (Environmental Protection Agency.) (1994) Water quality standards handbook: 2nd ed.
EPA/823/B-94/005a,b. Washington, DC: Office of Water. Available online on
http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=20003QXV.TXT.
U.S. EPA (Environmental Protection Agency). (1998a) Guidelines for ecological risk assessment.
EPA/630/R-95/002F. Washington DC: Risk Assessment Forum. Available online at
http://www2.epa.gov/osa/guidelines-ecological-risk-assessment.
U.S. EPA (Environmental Protection Agency). (1998b) Environmental Monitoring and Assessment Program -
surface waters: field operations and methods for measuring the ecological conditions of wadeable streams.
EPA/620/R-94/004F. Research Triangle Park, NC: National Exposure Research Laboratory, National Health and
Environmental Effects Research Laboratory. Available online at
https://archive.epa.gov/emap/archive-emap/web/html/ws abs.html.
U.S. EPA (Environmental Protection Agency). (2000a) Nutrient criteria technical guidance manual: Rivers and
streams. EPA/822/B-00/002. Washington, DC: Office of Water. Available online at
https://www.epa.gov/nutrient-policv-data/criteria-development-guidance-rivers-and-streams.
U.S. EPA (Environmental Protection Agency). (2000b) Stressor identification guidance document.
EPA/822/B-00/025. Washington, DC: Office of Water. Available online at
https ://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=20685.
U.S. EPA (Environmental Protection Agency). (2001) Laboratory methods for processing macroinvertebrate
samples, macroinvertebrate indicator standard operating procedures (SOP#701). Cincinnati, OH: National
Environmental Research Laboratory.
U.S. EPA (Environmental Protection Agency). (2002) Summary of biological assessment programs and biocriteria
development for states, tribes, territories, and interstate commissions: streams and wadeable rivers.
EPA/822/R-02/048. Washington, DC: Office of Water. Available at
http://water.epa.gOv/scitech/swguidance/standards/criteria/aalife/biocriteria/upload/2002-Summarv-BioAssess-
Programs.pdf.
U.S. EPA (Environmental Protection Agency). (2003) Generic ecological assessment endpoints (GEAEs) for
ecological risk assessment. EPA/630/P-02/004F. Washington, DC: Risk Assessment Forum. Available online
at https://www.epa.gov/sites/production/files/2014-l 1/documents/generic endpoinsts 2004.pdf.
8-15

-------
U.S. EPA (Environmental Protection Agency). (2004a) Wadeable streams assessment: Quality assurance project
plan. EPA/841/B-04/005. Washington, DC: U.S. Environmental Protection Agency, Office of Water and Office of
Research and Development. Available online at http://www.epa. gov/o wo w/monitoring/wsa/OAPP-August 18.pdf.
U.S. EPA (Environmental Protection Agency). (2004b) Wadeable streams assessment: Water chemistry laboratory
manual. EPA/84l/B-04/008. Washington, DC: U.S. Environmental Protection Agency, Office of Water and Office
of Research and Development. Available online at
http://www.epa.gov/owow/monitoring/wsa/WRS lab manual.pdf.
U.S. EPA (Environmental Protection Agency). (2006) Wadeable streams assessment: A collaborative survey of the
nation's streams. EPA-841-B-06-002. Washington, DC: Office of Research and Development. Office of Water.
Available online at: http://water.epa.gov/tvpe/rsl/monitoring/streamsurvev/
upload/2007 5 16 streamsurvev WSA Assessment Mav2007.pdf
U.S. EPA (Environmental Protection Agency). (2009) National rivers and streams assessment: field operations
manual. DC; EPA/84 l/B-07/009. Washington, DC: Office of Water. Available online at
http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=P100AVQF.txt.
U.S. EPA (Environmental Protection Agency). (2010) Primary distinguishing characteristics of Level III Ecoregions
of the Continental United States. Available online at:
ftp://ftp.epa.gov/wed/ecoregions/us/Eco Level III descriptions.doc.
U.S. EPA (Environmental Protection Agency). (201 la) A field-based aquatic life benchmark for conductivity in
Central Appalachian streams. EPA/600/R-10/023F. Washington, DC: Office of Research and Development,
National Center for Environmental Assessment. Available online at
http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=23 3 809.
U.S. EPA (Environmental Protection Agency). (201 lb) The effects of mountaintop mines and valley fills on aquatic
ecosystems of the Central Appalachian coalfields. EPA/600/R-09/138F. Washingon, DC: Office of Research and
Development, National Center for Environmental Assessment. Available online at
http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=P100E3KG.txt.
U.S. EPA (Environmental Protection Agency). (201 lc) Review of field-based aquatic life benchmark for
conductivity in Central Appalachian streams. Washington, DC: Science Advisory Board, Office of the
Administrator. Available online at:
http://vosemite.epa.gOv/sab/sabproduct.nsf/0/EEDF20B88AD4C6388525785E00733 lF3/$File/EPA-SAB-l 1-006-
unsigned.pdf.
U.S. EPA (Environmental Protection Agency). (201 Id). A primer on using biological assessments to support water
quality management. Office of Science and Technology Office of Water, Washington, DC EPA 810-R-11-01.
Available online at
http://citeseerx.ist.psu.edu/viewdoc/do wnload:isessionid=CD37911D739A54D92C380DB72D3B72CA?doi=10.1.1.
644.3913&rep=rep 1 &tvpe=pdf
U.S. EPA (Environmental Protection Agency). (2013a) Biological assessment program review: Assessing level of
technical rigor to support water quality management. EPA/820/R-13/001. Washington, DC: Office of Science and
Technology. Available online at:
http://water.epa.gov/scitech/swguidance/standards/criteria/aqlife/biocriteria/upload/2013biological assessmentpdf.
U.S. EPA (Environmental Protection Agency). (2013b) National rivers and streams assessment 2008-2009: A
collaborative survey. Draft. EPA/841/D-13/001. Washington, DC: Office of Wetlands, Oceans, and Watersheds.
Office of Research and Development. Available online at:
http://water.epa.gOv/tvpe/rsl/monitoring/riverssurvev/upload/NRSA0809 Report Final 508Compliant 130228.pdf
Data available online at: http://water.epa.gov/tvpe/rsl/monitoring/riverssurvev/index.cfm.
U.S. EPA (Environmental Protection Agency). (2013c) EMAP Data. Washington, DC: Environmental Monitoring &
Assessment Program. Available online at: http://www.epa. gov/emap/html/data/index.html.
8-16

-------
U.S. EPA (Environmental Protection Agency). (2013d) US Level III Ecoregions of the Continental United States.
Available online at https://www.epa.gov/eco-researcMevel-iii-and-iv-ecoregions-continental-united-states.
USGS (United States Geological Survey). (2016). Mount Rainier. Washington, Pierce County: Volcano Hazards
Program. Available online at: http://volcanoes.usgs.gov/volcanoes/mount rainier/.
Vafai, SB; Mootha, VK. (2012) Mitochondrial disorders as windows into an ancient organelle. Nature
491(7424):374-383. doi:10.1038/naturell707.
van Dam, RA; Hogan, AC; McCullough, CD; Houston, MA; Humphrey, CL; Harford, AJ. (2010) Aquatic toxicity
of magnesium sulfate, and the influence of calcium, in very low ionic concentration water. Environ Toxicol Chem
29:410-421.
Vander Laan, JJ; Hawkins, CP; Olson, JR; Hill, RA. (2013) Linking land use, in-stream stressors, and biological
condition to infer causes of regional ecological impairment in streams. Freshw Sci 32(3):801-820.
Veil, JA; Puder, MG; Elcock, D; Redweik, RJ, Jr. (2004) A white paper describing produced water from production
of crude oil, natural gas, and coal bed methane. Argonne, IL: U.S. Department of Energy, Argonne National
Laboratory. Available on-line at:
http://citeseerx.ist.psu.edu/viewdoc/download?rep=repl&tvpe=pdf&doi=10.1.1.211.8898.
Vesper, DJ; Moore, JE; Adams, JP. (2016) Inorganic carbon dynamics and CO2 flux associated with coal-mine
drainage sites in Blythedale PA and Lambert WV, USA. Environ Earth Sci 75(4): 1-4.
Wallace, JB; Merritt, RW. (1980) Filter-feeding ecology of aquatic insects. Ann Rev Entomol 25:103-132.
Wallace, H; Wallace, BMN; Badaway, GMI. (1997) Lampbrush chromosomes and chiasmata of sex-reversed newts.
Chronosoma 106:526-533.
Wallace, JB. (1990) Recovery of lotic macroinvertebrate communities from disturbance. Environ Manage
14(5):605-620.
Wallace, JB; Webster, JR. (1996) The role of macroinvertebrates in stream ecosystem function. Ann Rev Entomol
41:115-39.
Wang, N; Ingersoll, CG; Kunz, JL; Brumbaugh, WG; Kane, CM; Evans, RB; Alexander, S; Walker, C; Bakaletz, S.
(2013) Toxicity of sediments potentially contaminated by coal mining and natural gas extraction to unionid mussels
and commonly tested benthic invertebrates. Environ Toxicol Chem 32(1):207-221.
Wang, N; Dorman, RA; Ingersoll, CG; Hardest, DK; Brumbaugh, WG; Hammer, EJ; Bauer, CR; Mount, DR;
(2016a) Acute and chronic toxicity of sodium sulfate to four freshwater organisms in water-only exposures. Environ
Toxicol Chem 35:115-127.
Wang, N; Ivey, CD; Ingersoll, CG; Brumbaugh, WG; Alvarez, D; Hammer, EJ,;Bauer, CR; Augspurger, T;
Raimondo, S; Barnhart, MC. (2016b). Acute sensitivity of abroad range of freshwater mussels to chemicals with
different modes of toxic action. Environ Toxicol Chem doi: 10.1002/etc.3642. [Epub ahead of print].
Waters, TF. (1966) Production rate, population density, and drift of a stream invertebrate. Ecology 47:595-604.
Waters, TF. (1972) The drift of stream insects. Ann Rev Entomol 17:253-267.
Waters, TF. (1995) Sediment in streams: sources, biological effects and control. Bethesda, Maryland: American
Fisheries Society. 251 pp.
WDE (Washington State Department of Ecology. (2014) Standard Operating Procedures (SOPs) for sampling,
auditing, and field methodology. Environmental Assessment Program. Available online at:
http ://www. ecv. wa. gov/pro grams/eap/qualitv. html#freshwater.
8-17

-------
Wetzel, RG. (2001) Chapter 10. Salinity in inland waters. In: Limnology: lake and river ecosystems. 3rd Edition. San
Diego, CA: Academic Press. 1006 pp.
Werner, AD; CT. Simmons. (2009) Impact of sea-level rise on sea water intrusion in coastal aquifers. Ground Water
47(2): 197-204.
Whittier, TR; Hughes, RM; Stoddard, JL; Lomnicky, GA; Peck, DV; Herlihy, AT. (2007a) A structured approach to
developing indices of biotic integrity: three examples from western USA streams and rivers. Trans Am Fish Soc
136:718-735.
Whittier, TR; Stoddard, JL; Larsen, DP; Herlihy, AT. (2007b) Selecting reference sites for stream biological
assessments: best professional judgment or objective criteria. Journal of the North American Benthological Society.
2007 Jun;26(2):349-60.
Wichard, W; Tsui, PTP; Kommick, H. (1973) Effect of different salinities on the coniform chloride cells of mayfly
larvae. J Insect Physiol 19:1825-1835.
Wiens, JJ. (2004) Speciation and ecology revisited: Phylogenetic niche conservatism and the origin of species.
Evolution 58:193-197.
Williams D.D. 1996. Environmental constraints in temporary fresh waters and their consequences for the insect
fauna. JN AmBenthol Soc 15:634-636.
Williams, DD; Hynes, HBN. (1976) The recolonization mechanisms of stream benthos. Oikos 27 (2):265-272.
Wood-Eggenschwiler, S; Barlocher, F. (1983) Aquatic hyphomycetes in sixteen streams in France, Germany and
Switzerland. Trans BritMycol Soc 81:371-379.
Wood, CM; Shuttleworth, TJ. (2008) Cellular and molecular approaches to fish ionic regulation. Vol. 14: Fish
Physiology. San Diego, CA: Academic Press, Inc.
Wood, PJ; Dykes, AP. (2002) The use of salt dilution gauging techniques: Ecological considerations and insights.
Water Res 36:3054-3062.
Wood-Eggenschwiler, S; Barlocher, F. (1983) Aquatic hyphomycetes in sixteen streams in France, Germany and
Switzerland. Trans BritMycol Soc 81:371-379.
Woods, AJ; Omernik, JM; Brown, DD. (1996) Level III and IV ecoregions of Pennsylvania and the Blue Ridge
Mountains, the central Appalachian Ridge and Valley, and the central Appalachians of Virginia, West Virginia, and
Maryland. EPA/600/R-96/077. Corvallis, OR: U.S. Environmental Protection Agency, National Health and
Environmental Effects Research Laboratory.
http://training.fws.gov/courses/csp/csp3200/resources/documents/epa region 3 eco desc.pdf.
Woods, AJ; Omernick, JM; Brown, DD. (1999) Level III and IV ecoregions of Delaware, Maryland, Pennsylvania,
Virginia, and West Virginia. Corvallis, OR: U.S. Environmental Protection Agency, National Health and
Environmental Effects Research Laboratory. 58 pp. Available online at:
https://extension.umd.edu/sites/default/files/ docs/programs/master-
gardeners/Natives/1999 Woods Omernik reg3 ecoregion descriptions.pdf.
Woods, AJ; Omernik, JM; Nesser, JA; Shelden, J; Comstock, JA; Azevedo, SH. (2002) Ecoregions of Montana, 2nd
edition (color poster with map, descriptive text, summary tables, and photographs) Map scale 1:1,500,000. U.S.
Enviromental Protection Agency. Available online at http://www.epa.gov/wed/pages/ecoregions/mt eco.htm.
Woodward, DF; Hansen, JA; Bergman, H; Little, E; DeLonay, AJ. (1995) Brown trout avoidance of metals in
waters characteristic of the Clark Fork River, Montana. Can. J. Fish. Aquat. Sci. 52: 2013-2017.
Woodward, DF; Goldstein, JN; Farag, A; Brumbaugh, WG. (1997) Cutthroat trout avoidance of metals and
conditions characteristic of a mining waste site: Couer d'Alene River, Idaho. Trans Am Fish Soc 126:699-706.
8-18

-------
Wright, IA; Davies, PJ; Findlay, SJ; Jonasson, OJ. (2011) A new type of water pollution: concrete drainage
infrastructure and geochemical contamination of urban waters. Mar Freshwr Res 62:1355-1361.
WVDEP (West Virginia Department of Environmental Protection). (2013) Watershed Assessment Branch 2013
standard operating procedures. Charleston, VA: West Virginia Department of Environmental Protection. 443 pp.
Available online at: http://www.dep.wv. gov/WWE/watershed/Pages/WBSOPs.aspx.
WVDEP (West Virginia Department of Environmental Protection). (2008a) West Virginia integrated water quality
monitoring and assessment report. Charleston, VA: West Virginia Department of Environmental Protection.
Available online at:
http://www.dep.wv.gov/WWE/watershed/IR/Documents/IR 2008 Documents/WV IR 2008 Report Only EPA A
pproved.pdf.
WVDEP (West Virginia Department of Environmental Protection). (2008b) 2008 Standard operating
procedures. Vol. 1. Charleston, WV: Watershed Assessment Branch, WVDEP.
WVDEP (West Virginia Department of Environmental Protection). (2006) Quality assurance project plan for
watershed branch monitoring activities. Charleston, WV: Division of Water and Waste Management, WVDEP.
Available online at http://www.dep.wv.gov/WWE/watershed/Documents/2005 06 SOP All.pdf.
Yasuno, M; Fukushima, S; Hasegawa, J; Shioyama, F; Hatakeyama, S. (1982) Changes in benthic fauna and flora
after application of temephos to a stream on Mt. Tsukuba. Hydrobiologia 89:205-14.
Zaiontz, C. (2014) Real statistics using excel. Available online at
http://www.real-statistics.com/regression/confidence-and-prediction-intervals/.
Zalizniak, L; Kefford, BJ; Nugegoda, D. (2006) Is all salinity the same? I. The effect of ionic compositions on the
salinity tolerance of five species of freshwater invertebrates. Mar Freshw Res. 57(l):75-82.
Zhao, Q; Jia, X; Xia, R; Lin, J; Zhang, Y. (2016) A field-based method to derive macroinvertebrate benchmark for
specific conductivity adapted for small data sets and demonstrated in the Hun-Tai River Basin, Northeast China.
EnvironPollut 216: 902-910. http://dx.doi.Org/10.1016/i.envpol.2016.06.065.
Zhichang, M; Huanting, S; Liu, JT; Eisma, D. (2001) Types of saltwater intrusion of the Changjiang Estuary. Sci
China B 44(Suppl): 105-157.
Ziegler, CR; Suter, GW, II; Kefford, BJ. (2010) Candidate cause: ionic strength. U.S. Environmental Protection
Agency, CADDIS (The causal analysis/diagnosis decision information system). Available online at
www.epa. gov/caddis.
Zhang, X; Wakamatsu, H. (2002) A new equivalent circuit different from the Hodgkin-Huxley model, and an
equation for the resting membrane potential of a cell. Artif Life Robot 6:140-148.
Zielinski, RA; Otton, JK; Johnson, CA. (2001) Sources of salinity near a coal mine spoil pile, north central
Colorado. J Environ Qual 30:1237-1248.
8-19

-------