EPA 843-R-19-001
National Rivers and Streams
Assessment 2013-2014
Technical Support Document
U.S. Environmental Protection Agency
Office of Water: Office of Wetlands, Oceans and
Watersheds
Office of Research and Development
Washington, DC 20460
December 2020
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
U.S. Environmental Protection Agency. 2020. National Rivers and Streams Assessment 2013-2014
Technical Support Document. EPA 843-R-19-001. Office of Water and Office of Research and
Development. Washington, D.C. https:/ /www.epa.gov/riationai-aqiiatic-resource-surveys/nrsa
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Contents
List of Figures 7
List of Tables 8
List of Acronyms 10
1 Introduction 11
1.1 Additional Resources for Survey Operations 11
2 Quality Assurance 12
2.1 Introduction 12
2.2 Survey Design 13
2.2.1 Statistical Design 13
2.2.2 Completeness 14
2.2.3 Comparability 14
2.3 Quality Assurance In Field Operations 14
2.3.1 Field Method PilotTesting 14
2.3.2 Training of Field Trainers and Assistance Visitors 15
2.3.3 Field Crew Training 15
2.3.4 Field Assistance Visits 15
2.3.5 Revisits of Selected Field Sites 16
2.3.6 Evaluation of Fish Identifications 16
2.4 Laboratory Quality Assurance And Quality Control 16
2.4.1 Basic Capabilities 16
2.4.2 Benthic Macroinvertebrate Identifications 17
2.4.3 Chemical Analyses 17
2.5 Data Management And Review 18
2.6 Main Report 19
2.7 Literature Cited 19
3 Selection of Probability Sites 21
3.1 Objectives 21
3.2 Target Population 21
3.3 Sample Frame 22
3.4 Survey Design 23
3.4.1 NRSA09 Design 23
3.4.2 NRSA14 Design 24
3.4.3 Stratification 24
3.4.4 Multi-Density Categories 24
3
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
3.4.5 Oversample and Site Replacement 25
3.4.6 State Designs 26
3.5 Evaluation Process 26
3.6 Implementation of the design 27
3.7 Statistical Analysis 27
3.8 Literature Cited 28
4 Selection of Sites to Establish Reference Conditions 29
4.1 Sources of Reference Sites 30
4.2 Chemical and Physical Screens 31
4.3 Geospatial Screens 32
4.4 Literature Cited 33
5 Benthic Macroinvertebrates 37
5.1 Overview 37
5.2 Data Preparation 38
5.2.1 Standardizing Counts 38
5.2.2 Operational Taxonomic Units 38
5.2.3 Autecological Characteristics 38
5.3 Multimetric Index Development 39
5.3.1 Regional Multimetric Development 39
5.3.2 Modeling of MMI Benchmarks 42
5.4 Predicted O/E Modeling 43
5.5 Literature Cited 45
6 Fish Assemblage 47
6.1 Background 47
6.1.1 Multimetric Indicator for NRSA 2008-09 47
6.1.2 Multimetric Indicator for NRSA 2013-14 47
6.1.3 Regionalization 48
6.2 Methods 48
6.2.1 Field methods 48
6.2.2 Counting, Taxonomy, and Autecology 48
6.3 Fish multimetric index development 51
6.3.1 Least-Disturbed Reference Sites for Fish 52
6.3.2 Candidate Metrics 52
6.3.3 Adjustment of Metric Response for Watershed Area 52
6.3.4 Selection of Final Candidate Metrics 54
4
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.3.5 Metric Scoring 54
6.3.6 Selection of Final Fish MMIs 55
6.4 Fish MMI performance 69
6.5 Sites with low fish abundance 77
6.6 Benchmarks for assigning ecological condition 77
6.7 Discussion 80
6.8 Literature cited 82
Appendix 6.A Comparison of model-based and traditional fish multimetric indices for NRSA 2008-
09 85
6.A.1 Responsiveness to Disturbance and Precision 85
6.A.2 Repeatability and Sensitivity 88
6.A.3 Correlation of Fish MMI Scores 88
6.A.4 Population Estimates 91
Appendix 6.B Candidate metrics considered for fish MMI development 93
7 Water Chemistry Analyses 100
7.1 Acidity And Salinity Benchmarks 100
7.2 Total Phosphorus And Total Nitrogen Benchmarks 100
7.3 Signal to Noise 101
7.4 Literature Cited 101
8 Physical Habitat Assessment 103
8.1 Methods 104
8.1.1 Physical Habitat Sampling and Data Processing 104
8.1.2 Quantifying the Precision of Physical Habitat Indicators 106
8.2 Physical Habitat Condition Indicators 107
8.2.1 Relative Bed Stability and Excess Fines 107
8.2.2 Riparian Vegetation 110
8.2.3 Instream Habitat Cover Complexity Ill
8.2.4 Riparian Human Disturbances 113
8.3 Estimating Reference Condition for Physical Habitat 113
8.3.1 Reference Site Screening and Anthropogenic Disturbance Classifications 113
8.3.2 Modeling Expected Reference Values of the Indicators 114
8.4 Response of the Physical Habitat Indicators to Human Disturbance 118
8.5 Literature Cited 120
Appendix 8.A 147
9 Human Health Fish Tissue Indicator 163
9.1 Field Fish Collection 163
5
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
9.1.1 Fish Tissue Fillets 163
9.1.2 Fish Tissue Plugs 163
9.2 Mercury Analysis And Human Health Fish Tissue Benchmarks 164
9.3 PCB Analysis And Human Health Fish Tissue Benchmarks 165
9.4 PFAS Analysis And Human Health Fish Tissue Benchmark 165
9.5 Calculation Of Human Health Fish Tissue Benchmarks 167
9.6 Literature Cited 169
10 Enterococci Indicator 170
10.1 Field Collection 170
10.2 Lab Methods 170
10.3 Application Of Thresholds 171
10.3.1 Calibration 171
10.3.2 Thresholds 171
10.4 Literature Cited 171
11 Microcystins 173
11.1 Field Methods 173
11.2 Microcystin Analysis and Application of Benchmarks 173
11.3 Literature Cited 173
12 From Analyses to Results 175
12.1 Extent And Risk Estimation And Assessment 175
12.1.1 Condition Classes 175
12.1.2 Estimating the Extent for Each Condition 176
12.1.3 Relative Extent 176
12.1.4 Relative Risk 177
12.1.5 Attributable Risk 178
12.2 Difference Analyses 178
12.2.1 Data Preparation 178
12.2.2 Analysis 179
12.3 Literature Cited 179
6
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
List of Figures
Figure 4.1 Examples of percent urban (A, 60%) and row crop (B, 72%) from NLCD 33
Figure 6.1 Aggregated Omernik ecoregions used to develop model-based fish MMIs for NRSA 2008-
09 50
Figure 6.2 Aggregated Omernik ecoregions used to develop traditional fish MMIs for NRSA 2013-14
51
Figure 6.3 Boxplots comparing regional fish MMI scores of least-disturbed sites to most-disturbed
sites 72
Figure 6.4 Regional fish MMI scores versus Strahler order category (least-disturbed sites) 74
Figure 6.5 Regional fish MMI scores versus fish sampling protocol (least-disturbed sites) 75
Figure 6.6 Regional fish MMI scores versus stream temperature class (least-disturbed sites)..76
Figure 6.7 Relationship between number of fish collected, reduced habitat volume, and small
watershed size at least-disturbed sites 79
Figure 8.1 Sample sites for NRSA 2008-09 and NRSA 2013-14 135
Figure 8.2. Riparian Disturbance (Wl_Hall) in combined NRSA 2008-09 and 2013-14 sample sites
136
Figure 8.3 Riparian Disturbance (Wl_Hall) in combined NRSA 2008-09 and 2013-14 sample sites 137
Figure 8.4 Log Relative Bed Stability (LRBS_use) and LoglO geometric mean bed surface substrate
diameter 138
Figure 8.5 Observed/Expected Relative Bed Stability (LOE_LRBS_use) in combined NRSA 2008-09
and 2013-14 sample sites 139
Figure 8.6 Riparian Vegetation Cover Complexity (LPt01 JKCMGW) in combined NRSA 2008-09
and 2013-14 sample sites 140
Figure 8.7 Observed/Expected Riparian Vegetation Cover Complexity (LOE_XCMGW_use) in
combined NRSA 2008-09 and 2013-14 sample sites 142
Figure 8.8 Instream Habitat Complexity (LPt01 _XFC_NAT) in combined NRSA 2008-09 and 2013-
14 sample sites 144
Figure 8.9 Observed/Expected Instream Habitat Complexity (LOE_XFC_NAT_use) in combined
NRSA 2008-09 and 2013-14 sample sites 145
7
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
List of Tables
Table 3.1 Expected sample size for NRSA09 Design 24
Table 3.2 Expected sample size for NRSA14 Design 25
Table 3.3 Oversample replacement categories 26
Table 3.4 Recommended Codes for Evaluating Sites 27
Table 3.5 Evaluation Status of Dropped Sites 27
Table 4.1 Macroinvertebrate reference sites available for use in the NRSA 31
Table 4.2 Criteria for eight chemical and physical habitat filters used to identify the candidate least
disturbed reference sites for benthic macroinvertebrate and fish indicators for each of the nine
aggregate ecoregions for NRSA 35
Table 4.3 Criteria for eight chemical and physical habitat filters used to identify the candidate most-
disturbed sites for each of the nine aggregate ecoregions for NRSA 36
Table 5.1 Six benthic community metrics, scoring direction, and floor and ceiling values used in
calculating the NRSA and WSA MMI in each of the nine aggregate ecoregions 41
Table 5.2 MMI-Disturbance Regression Model Statistics Used for Setting Benchmarks 43
Table 5.3 Benchmark Values for the Nine Regional Benthic MMIs 43
Table 5.4 Benthic Macroinvertebrate Predictive Models 45
Table 6.1 Criteria used to select least-disturbed sites for use in developing the regional NRSA fish
multimetric indices (MMIs) based on 2008-09 data 53
Table 6.2 Number of final candidate fish multimetric indices (MMIs) 56
Table 6.3 Regression equations for adjusting metrics for watershed area 56
Table 6.4 Performance information of metrics used to construct the fish MMI for the Coastal Plain.. 59
Table 6.5 Performance information of metrics used to construct the fish MMI for the Northern
Appalachians 60
Table 6.6 Performance information of metrics used to construct the fish MMI for the Northern Plains
62
Table 6.7 Performance information of metrics used to construct the fish MMI for the Southern
Appalachians 63
Table 6.8 Performance information of metrics used to construct the fish MMI for the Southern Plains
64
Table 6.9 Performance information of metrics used to construct the fish MMI for the Temperate
Plains 66
Table 6.10 Performance information of metrics used to construct the fish MMI for the Upper Midwest
67
Table 6.11 Performance information of metrics used to construct the fish MMI for the Western
Mountains 68
Table 6.12 Performance information of metrics used to construct the fish MMI for the Xeric West... 70
Table 6.13 Performance statistics for the nine regional fish MMIs 71
Table 6.14 Determining the minimum drainage area expected to reliably support the presence of fish
(adapted from McCormick et al (2001)). Variable names are from the NRSA database. Scores
for each metric between the upper and lower criteria were estimated by linear interpolation. ..78
Table 6.15 Benchmarks for assigning ecological condition based on the distribution of regional fish
MMI scores 81
Table 7.1 Nutrient and Salinity Category Benchmarks for NRSA Assessment 101
Table 8.1 Metrics used to characterize the general attributes of stream/river physical habitat 126
Table 8.2 Sampling revisit precision (repeatability) of the four physical habitat condition indicators.. 127
Table 8.3 Estimated number of years to detect trends in habitat attributes 128
8
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.4 Anthropogenic disturbance screening criteria 129
Table 8.5 NRSA boatable and wadeable least-disturbed reference sites from combined 2008-09 &
2013-14 surveys 130
Table 8.6 Summary of regression models used in estimating site-specific expected values of LoglO
Relative Bed Stability (LRBS^OS) under least-disturbed reference conditions 131
Table 8.7 Summary of regression models used in estimating site-specific expected values of Riparian
Vegetation Cover and Structure (LoglO[0.01+XCMGW]) under least-disturbed reference
conditions 132
Table 8.8 Summary of regression models used in estimating site-specific expected values of Instream
Habitat Cover Complexity (LoglO[0.01+XFC_NAT]) under least-disturbed reference
conditions 133
Table 8.9 Responsiveness to levels of human disturbance 134
Table 9.1 Recommended Target Species for Fish Tissue Indicator Sample Collection 164
Table 9.2 NRSA 2013-14 Composite Fish Fillet Tissue Summary Data 167
Table 12.1 Simplified Notation 177
9
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
List of Acronyms
ANC
Acid Neutralizing Capacity
CCE
Calibrator Cell Equivalent
CPL
Coastal Plain ecoregion
DII
Dam Influence Index
DOC
Dissolved Organic Carbon
EMAP
EPA's Environmental Monitoring and Assessment Program
EPA
U.S. Environmental Protection Agency
FFG
Functional feeding group
FMMI
Fish Multimetric Index
HUC
Hydrologic Unit Codes
IBI
Index of Biotic Integrity
IQR
Interquartile Range
Km
kilometers
MAHA
Mid-Atlantic Highlands Assessment
MAIA
Mid-Atlantic Integrated Assessment
MMI
Multimetric Index
NAP
Northern Appalachians ecoregion
NARS
National Aquatic Resource Surveys
NAWQA
National Ambient Water Quality Assessment
NLCD
National Land Cover Dataset
NPL
Northern Plains ecoregion
NRSA
National Rivers and Streams Assessment
O/E
Ratio of Observed to Expected
OTU
Operational Taxonomic Unit
PCA
Principal Component Analysis
QA
Quality Assurance
QC
Quality Control
RBS
Relative Bed Stability
RF
Random Forest
RMSE
Root Mean Squared Error
S:N
Signal to Noise ratio
SAP
Southern Appalachians ecoregion
SD
Standard Deviation
SPL
Southern Plains ecoregion
TPL
Temperate Plains ecoregion
UMW
Upper Midwest ecoregion
WMT
Western Mountains ecoregion
WSA
Wadeable Streams Assessment 2000-2004
XER
Xeric ecoregion
10
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
1 Introduction
National divers and Streams Assessment 2013-2014: The Second Collaborative Survey is the second in a
series of National Rivers and Streams Assessment (NRSA) reports that utilize a randomized
statistical survey design to assess the quality of the nation's perennial rivers and streams. The
NRSA is one of the National Aquatic Resource Surveys (NARS), a set of collaborative programs
between EPA, states, and tribes designed to assess the quality of the nation's waters using a
statistical survey design. The survey data underlying this NRSA report were collected in the
summers of 2013 and 2014; as such, the findings presented in the report show a snapshot in time.
The key goals of the NRSA report are to describe the ecological and recreational quality of the
nation's perennial river and stream resources, how those conditions are changing, and the key
stressors affecting those waters. Clean Water Act (CWA) Sections 104(a) and (b) collectively grant
the Administrator authority to investigate and report on water quality across the country. NARS
data also inform and benefit the national water quality inventory report that EPA prepares for
Congress pursuant to CWA Section 305(b)(2).
This technical support document provides information about the analytical approaches used for
the NRSA 2013-14. National results from NRSA are included in the National Rivers and Streams
Assessment 2013-2014: The Second Collaborative Survey report and results for subpopulations, including
EPA regions and ecological regions, are presented in the online data dashboard
(https://rivers tTeamassessment-.epa.gov/dash board).
1.1 Additional Resources for Survey Operations
A series of protocols were used to ensure consistency throughout the survey operations. The
following documents provide the field sampling methods, laboratory procedures, quality measures,
and site selection for the NRSA 2013-14. Data from the survey are available to download at
https:/ /www.epa.gov/ national-aauatic-resouree-surveys / data-riational-aquatic-resource-surveys.
U.S. EPA. 2013. National Rivers and Streams Assessment: Field Operations Manual. EPA-
841-B-12-009a and EPA-841-B-12-009b. Washington, D.C.
U.S. EPA. 2014. National Rivers and Streams Assessment: Laboratory Operations Methods
Manual. EPA 841-B-12-010. Washington, D.C.
U.S. EPA. 2015. National Rivers and Streams Assessment: Quality Assurance Project Plan.
EPA 841-B-12-007. Washington, D.C.
U.S. EPA. 2012. National Rivers and Streams Assessment: Site Evaluation Guidelines. EPA
841-B-12-008. Washington, D.C.
11
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
2 Quality Assurance
The NRSA implemented and assessed the quality of its operations and data throughout the NRSA
2013-14 survey. This chapter documents the NRSA's adherence to the requirements of EPA's quality
system implemented by the Office of Water as explained in the introduction section below. The
following sections describe the quality aspects of the statistical design, field operations, laboratory
assessments, data management, and report writing.
2.1 Introduction
The EPA quality system incorporates a national consensus standard for quality systems authorized by
the American National Standards Institute (ANSI) and developed by the American Society for
Quality Control (ASQC), ANSI/ASQC E4-2004, Quality Systems for 'Environmental Data and Technology
Programs Requirements with Guidance for Use. EPA Order CIO 2105.0, dated May 5, 2000, requires all
of its component organizations to participate in an agency-wide quality system. The EPA Order also
requires quality assurance project plans or "equivalent documents" for all projects and tasks involving
environmental data.
In accordance with the EPA Order, the Office of Water (OW) developed the Office of Water
Quality Management Plan (QMP; USEPA 2015) to describe OW's quality system that applies to all
water programs and activities, including the NRSA, collecting or using environmental data. As
required by the EPA Order and OW QMP, the NRSA developed and abided by its QAPP
throughout the survey. One significant challenge encountered was application of the quality control
procedures for periphyton. As a result, EPA did not include periphyton in the NRSA 2013-14 report
and is working with the United States Geological Survey (USGS) and other experts to improve
periphyton (specifically diatom) taxonomy through development of tools and training materials. The
NRSA QAPP contains elements of the overall project management, data quality objectives,
measurement and data acquisition, and information management. The QAPP also deals with the data
integration necessary between the Wadeable Streams Assessment (WSA), the NRSA, and EPA's
Environmental Monitoring and Assessment Program (EMAP) Western Pilot Study (2001-2004) to
create one comprehensive report on the ecological status of the nation's rivers and streams.
The following companion documents to the QAPP present detailed procedures for implementing
the field and lab work for the NRSA 2013-14 survey:
National Rivers and Streams Assessment: Site Evaluation Guidelines EPA 841-B-12-008
National Rivers and Streams Assessment: Field Operations Manual (FOM), EPA-841-B-12-
009a and EPA-841-B-12-009b
National Rivers and Streams Assessment: Laboratory Operations Manual (LOM), EPA 841-
B-12-010
The four documents together address all aspects of the NRSA's data acquisition and evaluation. The
12
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
LOM also lists measurement quality objectives (MQOs) which were used to evaluate the level of
quality attainment for individual survey metrics. Every person involved in the NRSA was responsible
for abiding by the QAPP and adhering to the procedures specified in its companion documents.
Moreover, NRSA participants were instructed and/or trained in the requirements applicable to the
person's role in the survey {e.g., field crews were trained in the FOM procedures and applicable
QAPP requirements). For example, field crews attended a combined classroom and hands-on
training in field procedures. Laboratory personnel provided appropriate SOPs and certifications; and
attended calls to discuss implementation of the lab procedures.
2.2 Survey Design
The NRSA's survey design was based upon statistical concepts that are well accepted by the scientific
community. As described in the following sections, the survey design objectives were met by
requirements of the statistical design, completeness of implementing the design, and consistency with
established procedures.
2.2.1 Statistical Design
There is a large body of statistical literature dealing with sample survey designs which addresses the
challenge of making statements about many by sampling the few (Kish 1965). Sample surveys have
been used in a variety of fields (e.g., monthly labor estimates) to determine the status of populations
of interest, especially if the population is too numerous to census or if it is unnecessary to census the
population to reach the desired level of precision for describing the population's status. In natural
resource fields, probability sampling surveys have been consistently used to estimate the conditions
of the entire population. For example, the National Agricultural Statistics Survey (NASS) conducted
by the U.S. Department of Agriculture and the Forest Inventory Analysis (FIAT) conducted by the
U.S. Forest Service (Bickford et al. 1963, Hazard and Law 1989) have both used probability-based
sampling concepts to monitor and estimate the condition and productivity of agricultural and forest
resources from a commodity perspective. The sampling design strategy for NRSA is based on the
fundamental requirement for a probability sample of an explicitly defined resource population,
where the sample is constrained to reflect the spatial dispersion of the population. This design has
been documented in peer reviewed literature (Stevens 1994, Stevens and Olsen 1999). By applying
the statistical concepts of this design, the survey was able to meet the following overarching data
quality objectives:
In the conterminous U.S., estimate the proportion of perennial river and stream length (+ 5
percent) that falls below the designated benchmark for "good" conditions for selected
measures with 95 percent confidence.
For each of the aggregated Omernik Level III Ecoregions, estimate the proportion of
perennial river and stream length (ฑ15 percent) that falls below the designated benchmark
for "good" conditions for selected measures with 95 percent confidence.
13
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
2.2.2 Completeness
To ensure that the implementation of the NRSA 2013-14 sample design resulted in adequate
measurements, the survey included completeness requirements for field sampling and laboratory
analyses. The QAPP requires that valid data for individual indicators must be acquired from a
minimum number of sampling locations to make subpopulation estimates with a specified level of
confidence or sampling precision. As the starting place for selecting field sites, EPA used the
National Hydrography Dataset (NHD; https://www.usgs.gov/core-science-systems/no~p/national-
hydrography) as the frame representing streams and rivers in the US because it was the most
complete source of stream hydrology available at the national scale.1 The data completeness
requirements were achieved, and sites where data for an indicator could not be collected were
classified as "Not Assessed" in the population estimates.
2.2.3 Comparability
Comparability is defined as the confidence with which one data set can be compared to another
(Stanley and Verner, 1985; Smith et al, 1988). For all indicators, NRSA ensured comparability by the
use of standardized sampling procedures, sampling equipment, and analytical methodologies by all
sampling crews and laboratories. For all measurements, reporting units and format are specified,
incorporated into standardized data recording forms, and securely transferred into a centralized
information management system. Because NRSA 2013-14 used the same comparable methods to
collect data in EMAP West and WSA studies, the data also can be compared across the studies. The
following sections on field and laboratory operations describe additional measures to ensure
consistency in NRSA.
2.3 Quality Assurance In Field Operations
The requirements and methods presented in the Field Operations Manual (FOM) ensured that
quality objectives were attainable and survey activities were manageable. As described below,
NRSA tested its FOM, trained crews using the FOM, visited crews during the field season, and
confirmed fish specimen identifications.
2.3.1 Field Method Pilot Testing
Representatives from the NRSA team, logistics and data management contractors, and state partners
1 As EPA and the Department of the Army recognize in the Navigable Waters Protection Rule, "NHD at High Resolution
. . . may not accurately identify on-the-ground flow conditions." 85 FR 22294 (April 21, 2020). NHD-Plus maps surface
waters at a coarser resolution (1:100,000) compared to the scale of NHD at High Resolution (1:24,000). 4,566 sites were
evaluated as part of NRSA 2013-14. Of those, a total of 1,853 were sampled. 1,328 sites were target sites but not sampled
(landowner denial, otherwise inaccessible or other), and 1,385 sites were identified as non-target. 755 of the 1,385 non-
target sites were identified as non-perennial.
14
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
tested sampling methods, paper and electronic field forms, and equipment described in the FOM.
The test run assessed the accuracy and clarity of the FOM's instructions for executing the
procedures and quality steps. The test run also evaluated sampling logistics, sample preparation, and
sample shipping instructions. As a result of lessons learned during the test run, NRSA staff corrected
and improved the FOM prior to field crew training.
2.3.2 Training of Field Trainers and Assistance Visitors
Before training field crews, members of the NRSA team, oversight staff, contractor trainers, and
other experts tested the training materials during an intensive 4-day period that included classroom
and hands on training sessions. This "train-the-trainer" event serves two primary purposes. First,
the event is designed to make sure that all trainers understand the methods and are providing
consistent instruction to field crews. Second, it provides another opportunity to ensure that the
field documents and forms are clear and accurate. During this training event, the attendees tested
the materials to ensure that the instructions were correct and easy to execute and practiced actually
training the methods. The training materials included the FOM, Quick Reference Guide (QRG),
field forms, and PowerPoint presentations. As a result of the training, practice training sessions and
expert discussions, NRSA staff corrected and improved training materials, and the FOM and QRG
before the field crew training.
2.3.3 Field Crew Training
To ensure consistency across field crews, all field crews were required to attend a 4-day training
session prior to visiting any field site. At a minimum, the field crew leader and the fish taxonomist
from each crew were required to attend. NRSA trainers led regional field crew training sessions
consisting of classroom and field-based lessons. The lessons included sessions on conducting site
reconnaissance, recording field observations and in situ data, collecting field samples, preparing,
packing and shipping sample containers, and use of the standardized field forms. The field crew
leaders were taught to review every form and verify that all hand-entered data were complete and
correct.
2.3 A Field Assistance Visits
To further assist the crews in correctly implementing the field procedures and quality steps, a NARS
staff member or contractor trainer visited every NRSA field crew during the field season. These
visits, known as assistance visits (AV), provided an opportunity to observe field crews in the normal
course of a field day, assist in correctly applying the procedures, and document the crew's adherence
to sampling procedures. 223 AVs were completed in the summers of 2013 and 2014. If
circumstances were noted where a field crew was not conducting a procedure properly, the observer
recorded the deficiency, reviewed the appropriate procedure with field team, and assisted the field
crew until the procedure was completed correctly.
15
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
2.3.5 Revisits of Selected Field Sites
To evaluate within-year sampling variability, the NRSA design called for crews to revisit 10 percent of
the sites. These sites were sampled twice in the NRSA index period during a single year (visit 1 and
visit 2). Useful metrics and indicators tend to have high repeatability, that is among site variability will
be greater than sampling variability based on repeat sampling at a subset of sites. To quantify
repeatability, NARS uses Signal:Noise (S:N), or the ratio of variance associated with sampling site
(signal) to the variance associated with repeated visits to the same site (noise) (Kaufmann et al. 1999).
All sites are included in the signal, whereas only revisit sites contribute to the noise component.
Metrics with high S:N are more likely to show consistent responses to human caused disturbance, and
S:N values < 1 indicate that sampling a site twice yields as much or more metric variability as
sampling two different sites (Stoddard et al. 2008). The S:N values were used by analysts in the
process of selecting metrics and evaluating indicators.
2.3.6 Evaluation of Fish Iden tifications
To ensure consistent naming conventions, field taxonomist and laboratory ichthyologists were
required to use commonly accepted taxonomic references to identify fish vouchers. To evaluate their
identifications, field taxonomists were required to send fish vouchers from one or more site visits to
expert ichthyologists for a second, independent identification. The ichthyologists were able to
determine the taxa for 2,481 vouchers which was -10 percent of the 26,030 unique fish by site visit
collected for NRSA 2013-14. Overall, 82.3 percent of the 1,239 comparable 2013 records agreed at
the species level, 93.7 percent agreed at the genus level, 98.3 percent at the family level, and 99.0
percent at the order level. For 2014, 85.6% of the 1,242 comparable 2014 records agreed at the
species level, 95.9 percent agreed at the genus level, 99.4 percent at the family level, and 99.6 percent
at the order level.
2.4 Laboratory Quality Assurance And Quality Control
The NRSA laboratories used standard methods and/or followed the requirements (e.g.,
performance-based objectives) in the Laboratory Operations Manual (LOM). The QAPP identified
the overall quality requirements and the LOM provided methods that could be used to achieve the
quality requirements. If a laboratory chose a different method, it still had to meet the QA
requirements as described below.
2.4.1 Basic Capabilities
All laboratories were required to submit documentation of their analytical capabilities prior to
analyzing any NRSA 2013-14 sample. NRSA team members reviewed documentation to ensure that
the laboratories could meet required measurement quality objectives (MQOs; e.g., reporting limits,
detection limits). National Environmental Laboratory Accreditation Conference (NELAC)
certification, satisfactory participation in round-robin, or other usual and customary types of
16
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
evaluations were considered acceptable capabilities documentation.
2A.2 Benthic MacroinvertebrateIdentifications
For benthic macroinvertebrate taxonomy, laboratories were required to use the same taxa lists,
conduct regular internal QC checks, and participate in an independent quality check. All
participating laboratories identified organisms using the most appropriate technical literature that
was accepted by the taxonomic discipline and reflected the accepted nomenclature at the time of the
survey. USGS BioData (https:/ /aaiiatic.biodata.iisgs.gov/) and the Integrated Taxonomic
Information System (ITIS. https://www.itis.gov/) were also used to verify nomenclatural validity
and reporting.
Taxonomic accuracy is evaluated by comparing identifications of the same organisms by primary
and secondary, independent laboratories. Each primary laboratory provided organisms from 10
percent of its samples, or least three samples if they had less than 10, to a secondary laboratory
for an independent evaluation. EPA, supported by an expert contractor, assessed the primary and
secondary identifications and then held reconciliation calls to allow the taxonomists to discuss
organisms that were identified differently. As part of this process, recommendations and
corrective actions were identified to address inaccurate taxonomic identification; and
measurement objectives were calculated to ensure the data were of sufficient quality for the
NRSA.
Of the 2,256 benthic macroinvertebrate samples, the secondary laboratory identified organisms in
202 samples. The mean percent taxonomic disagreement (PTD) between laboratories was 12
percent (better than the NRSA measurement objective of 15 percent as identified in the QAPP).
The overall percent difference in enumeration (PDE) was 2 percent (better than the NRSA
measurement objective of 5 percent as identified in the QAPP).
Even when the measurement objectives were met, laboratories implemented recommendations and
corrective steps for the QC samples and all other samples with the same organisms. If, for example,
it was evident that empty mollusk shells were being identified and recorded in one or more of the
QC samples, the laboratories needed to verify that they had not counted empty mollusk shells in
their other samples.
2.4.3 Chemical Analyses
For quality assurance of chemical analyses, laboratories used QC samples which are similar in
composition to samples being measured. They provide estimates of precision and bias that are
applicable to sample measurements. To ensure the ongoing quality of data during analyses, every
batch of water samples was required to include QA samples to verify the precision and accuracy of
the equipment, reagent quality, and other quality measures. These checks were completed by
analyzing blanks or samples spiked with known or unknown quantities of reference materials,
17
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
duplicate analyses of the same samples, blank analyses, or other appropriate evaluations. The
laboratories reported QA results along with each batch of sample results. In addition, laboratories
reported holding times. Holding time requirements for analyses ensure analytical results are
representative of conditions at the time of sampling. The NARS team reviewed the data and noted
any quality failures. The data analysts used the information about quality to determine whether to
include or exclude data in the evaluations. As described in the next section, the consolidated NRSA
database was further evaluated for quality issues.
2.5 Data Management And Review
Information management (IM) is integral to all aspects of the NRSA from initial selection of
sampling sites through dissemination and reporting of final, validated data. Quality measures
implemented for the IM system are aimed at preventing corruption of data at the time of their initial
incorporation into the system and maintaining the integrity of data and information after
incorporation into the system.
Reconnaissance, field observation and laboratory analysis data were transferred from NRSA survey
participants and collected and managed by the NARS IM center. Data and information were
managed using a tiered approach. First, all data transferred from a field team or laboratory were
physically organized {e.g., system folders) and stored in their original state. Next, NARS IM created a
synthesized and standardized version of the data to populate a database that represented the primary
source for all subsequent data requests, uses and needs. All samples were tracked from collection to
the laboratory.
The IM staff applied an iterative process in reviewing the database for completeness, transcription
errors, formatting compatibility, consistency issues and other quality control-related topics. This
first-line data review was performed primarily by NARS IM in consultation with the NRSA QA
team. A second-phase data quality review consisted of evaluating the quality of data based onMQOs
as described in the QAPP. This QA review was performed by the NRSA QA team using a variety of
qualitative and quantitative analytical and visualization approaches. Data that met the MQOs were
used without restriction. Data that did not meet the MQOs were qualified and further evaluated to
determine the extent to which quality control results deviated from the target MQOs. Minor
deviations, such as the field latitude and longitude did not fall on the mapped flow line, were noted
and qualified but did not prevent data from being used in analyses. Major deviations were also noted
and qualified, but data were excluded from the analyses. An example of a major deviation was
insufficient fish assemblage sampling; when this occurred, the fish multimetric index was not
calculated for a given site. Data not used for analyses because of quality control concerns account
for a subset of the missing data for each indicator analysis and add to the uncertainty in condition
estimates.
18
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
2.6 Main Report
The main report provides a summary of the findings of each of the data analyses and EPA's
interpretation of them. After the main report was extensively reviewed in-house by the NRSA team,
its partners, and other EPA experts, the report underwent outside peer review. The outside review
was the final step in ensuring that the main report and its findings met the quality requirements of
the QAPP. EPA contracted with an outside firm to conduct an Independent External Peer Review
(IEPR) of the main report. The firm selected three peer reviewers who were experts in water
resource monitoring and biological and ecosystem assessments. The firm provided the reviewers
with a copy of the main report, along with supporting documentation and a charge that solicited
comments specifically on the technical content, completeness and clarity, and scientific integrity of
the main report. EPA used the comments from the peer reviewers to refine and review the main
report.
2.7 Literature Cited
American National Standards Institute and American Society for Quality Control (ANSI/ASQC).
2004. Quality Systems for 'Environmental Data Collection and Environmental Technology Programs:
Collection and Evaluation of Environmental Data. E4-2004. Milwaukee, WI.
Bickford, C.A., C.E. Mayer, and K.D. Water. 1963. An Efficient Sampling Design for Forest
Inventory: The Northeast Forest Resurvey. Journal of Forestry. 61: 826-833.
Kaufmann, P.R., P. Levine, E.G. Robison, C. Seeliger, and D.V. Peck. 1999. Quantifying Physical
Habitat in Wadable Streams. EPA/620/R_99/003. U.S. Environmental Protection Agency,
Washington, DC.
Kish, L. 1965. Survey Sampling. John Wiley & Sons. New York. 643 pp.
Smith, F., S. Kulkarni, L.E. Myers, and M.J. Messner. 1988. Evaluating and presenting quality
assurance data. Pages 157-68 in L.H. Keith, ed. ACS Professional Reference Book. Principles of
Environmental Sampling. American Chemical Society, Washington, D.C.
Stanley, T.W., and S.S. Verner. 1986. The U.S. Environmental Protections Agency's quality assurance
program, pp. 12-19 In: J.K. Taylor and T.W. Stanley (Eds). Quality Assurance for Environmental
Measurements. ASTM STP 867, American Society for Testing and Materials, Philadelphia,
Pennsylvania.
Stevens Jr., D.L. 1994. Implementation of a National Monitoring Program. Journal Environmental
Management 42:1 -29.
Stevens Jr., D.L., and A.R. Olsen. 1999. Spatially restricted surveys over time for aquatic resources.
Journal of Agricultural, Biological, and Environmental Statistics 4:415-428.
Stoddard, J.L., A.T. Herlihy, D.V. Peck, R.M. Hughes, T.R. Whittier, and E. Tarquinio. 2008 A
process for creating multimetric indices for large-scale aquatic surveys. Journal of North American
Benthological Society 27: 878-891.
Stribling, J.B., S.R. Moulton, and G.T. Lester. 2003. Determining the quality of taxonomic data. Journal
of the North American Benthological Society 22(4) :621 -631.
19
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
USEPA. March 2015. Office of Water Quality Management Plan. EPA EPA-821-F-15-000. Revision 4.0.
U.S. Environmental Protection Agency, Office of Water. Washington, DC.
USEPA. February 2009. National divers and Streams Assessment: Site Evaluation Guidelines. EPA-841-B-
07-008. U.S. Environmental Protection Agency, Washington, DC.
USEPA. April 2009. National divers and Streams Assessment: Yield Operations Manual. EPA-841-B-07-9
U.S. Environmental Protection Agency, Washington, DC.
USEPA. November 2009. National divers and Streams Assessment: luiboratoiy Methods Manual. EPA-841-
B-07-010. U.S. Environmental Protection Agency, Washington, DC.
USEPA. December 2010. National divers and Streams Assessment: Integrated Quality Assurance Project Plan,
Final Document. EPA-841-B-07-007. U.S. Environmental Protection Agency, Office of Water
and Office of Research and Development, Washington, DC.
USEPA. May 2000. Order CIO 2105.0, Policy and Program Requirements for the Mandatory Agency-
wide Quality System. U.S. Environmental Protection Agency, Washington, DC.
20
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
3 Selection of Probability Sites
Using a statistical survey design, 1,853 sites were selected at random to represent the quality of the
larger population (1.2 million miles) of perennial rivers and streams across the lower 48 states, from
large rivers to small headwater streams. Sites were selected using a random sampling technique that
uses a probability-based design described in this chapter. The following sections describe the
statistical objectives, target population, sample frame, survey design, evaluation, and statistical
analysis.
3.1 Objectives
The statistical design requirements for NRSA 2013-14 were:
to estimate the proportion of perennial rivers and streams with a margin of error of + 5% in the
conterminous U.S. that fall below the designated benchmark for good conditions for selected
indicators with 95% confidence
to estimate the proportion of perennial rivers and streams with a margin of error of + 15% in
each of nine ecological reporting regions that fall below the designated benchmark for good
conditions for selected indicators with 95% confidence.
to estimate the difference in proportion of perennial river and streams in the conterminous U.S.
from 2008-09 to 2013-14 that fall below the designated benchmark for good (or poor) condition
for selected indicators. Difference estimates should have a margin of error of + 15% at 95%
confidence.
to estimate the difference in proportion of perennial river and streams in the conterminous U.S.
from 2008-09 to 2013-14 in each of nine ecological reporting regions that fall below the
designated benchmark for good (or poor) condition for selected indicators. Difference estimates
should have a margin of error of + 15% at 95% confidence.
accomplish the above while ensuring that the minimum sample size for a state will be 20.
revisit 10% of the sites in 2013-14 for variance component estimation and quality assurance.
3.2 Target Population
The target population consisted of all streams and rivers within the 48 contiguous states that
had flowing water during the study index period (i.e., beginning of June through end of September for
most regions). This included major rivers and small streams. Sites must have had > 50% of the reach
length with standing water and sites were to be sampled during base flow conditions. Sites with water
in less than 50% of the reach length were dropped. The target population excludes tidal rivers and
streams up to head of salt (defined as < 0.5 ppt for this study), as well as run-of-the-river ponds and
reservoirs with greater than 7-day residence time.
21
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
3.3 Sample Frame
The sample frame, used to represent the target population, was derived from the medium resolution
National Hydrography Dataset (NHD-Plus). Attributes from NHD-Plus and additional attributes
added to the sample frame that are used in the survey design are:
MajorRiver: rivers identified as major rivers or additional rivers in the book: Rivers of North
America (Benke and Cushing 2005)
Strahler order
Strahler category where categories are RiversMajor (5th order and higher), RiversOther (5th order
and higher), LargeStreams (3rd,4th order), and SmallStreams (1st, 2nd order)
BorderRiver: rivers and streams that occur on state and country boundaries. Each reach is
identified by two-state postal codes such as MO:IL for the portion of the Mississippi River that
forms the boundary between Missouri and Illinois. A border river/stream is assigned to one of
the two states for the survey design
Ecological Reporting Region: Nine aggregated Omernik ecoregions that are used for reporting
Omernik and North American ecoregions Levels I, II, III and IV
Postal code (state)
Urban and non-urban rivers and streams
Landownership as non-federal, Forest Service, BLM, Tribal Land, US Fish and Wildlife Service,
US National Park Service, and Department of Defense
The Urban/non-urban attribute was created by intersecting a modified version of the Census Bureau
national urban boundary GIS coverage with NHD-Plus. The Census Bureau's boundaries were
buffered 100 meters to include a majority of stream features intersecting and coincident with urban
areas. Where this buffer did not completely gather all the river features within the urban areas (rivers
intersecting cities are excluded from the Census Bureau's urban areas), the NHD-Plus river area
(polygon) features were clipped at a three-kilometer buffer around the urban areas and combined
with the buffered urban area to create the modified urban database. If a stream or river segment was
within this boundary, it is designated as "Urban"; otherwise it is designated as "NonUrban".
FCODE is directly from NHD-Plus and is used to identify which segments in NHD were included in
the sample frame. The FCODEs are a numeric identifier of the channel type. The attribute Frame07
identifies each segment as either "Include" or "Exclude." Frame07 was created so that segments
included in the sample frame could be easily identified. All segments chosen to be sampled were
evaluated in the field prior to sampling to ensure they met the target population of NRSA (i.e.,
perennial rivers and streams). Sites that were not perennial were not sampled but were instead
replaced by the next perennial segment in the list. FCODE values included in the GIS shapefile:
FCODEs Included in 2013-2014 sample frame:
33600 Canal/Ditch
42801 Pipeline: Pipeline Type = Aqueduct; Relationship to Surface = At or Near
22
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
46000 Stream/River
46006 Stream/River (Perennial)
58000 Artificial Path (removed from dataset if coded through Lake/Pond and Reservoirs)
FCODEs Excluded in 2013-2014 sample frame
33400 Connector
46003 Stream/River (Intermittent)
42800 Pipeline
42802 Pipeline: Pipeline Type = Aqueduct; Relationship to Surface = Elevated
42803 Pipeline: Pipeline Type = Aqueduct; Relationship to Surface = Underground
42804 Pipeline: Pipeline Type = Aqueduct; Relationship to Surface = Underwater
42806 Pipeline: Pipeline Type = General Case; Relationship to Surface = Elevated
4280 Pipeline: Pipeline Type = General Case; Relationship to Surface = Underground
42809 Pipeline: Pipeline Type = Penstock; Relationship to Surface = At or Near
42811 Pipeline: Pipeline Type = Penstock; Relationship to Surface = Underground
42813 Pipeline: Pipeline Type = Siphon
56600 Coastline
58000 Artificial Path if coded through Lake/Pond and Reservoirs
3.4 Survey Design
The survey design consists of two major components (NRSA14 design and NRSA09 design) in order
to address the dual objectives of (1) estimating current status of perennial rivers and streams and (2)
estimating differences in status for perennial rivers and streams.
3.4.1 NRSA09 Design
The NRSA09 survey design is a subsample of the NRSA 2008-09 sites that were in the target
population and sampled in the NRSA 2008-09. The major objective for this design is difference
estimation, although all sites sampled in 2013-14 were used when differences are estimated.
The expected sample sizes were based on the nine ecological reporting regions and two Strahler order
categories of Rivers (5th and greater) and Streams (1st through 4th). Three ecological reporting regions
(UMW, NPL, SPL) involve a smaller number of states and were allocated fewer sites than the other
six regions (NAP, SAP, CPL, TPL, WMT, XER). Given these expected sample sizes, the number of
sites for each state was allocated proportional to the medium resolution NHD Plus perennial River or
Stream length in each state for each ecological reporting region.
23
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 3.1 Expected sample size for NRSA09 Design
Ecological
Region
Expected Sample Size NRSA09
Rivers
Stream
Total
NAP
45
50
95
SAP
45
50
95
CPL
45
50
95
TPL
45
50
95
XER
45
50
95
WMT
45
50
95
UMW
40
40
80
NPL
40
40
80
SPL
40
40
80
Total
390
420
810
The overall survey design included having 10% of the sites be visited twice in 2013-14. This was
accomplished by allocating four sites (two Rivers sites and two Streams sites) to each of the 48 states
for revisits (192 sites total). All of these revisit sites were assigned to the NRSA09 design. Moreover,
the sites selected to be revisited were also the same sites that were visited twice in 2008-09. This
results in 192 sites that were visited twice in 2008-09 and in 2013-14. The NRSA09 Design sites will
also be resampled in NRSA 2018-19.
3.4.2 NRSA14 Design
The NRSA14 survey design is a new survey design that selected new sites. The expected sample sizes
were based on the nine ecological reporting regions and four categories of RiversMajor (5th and
greater), RiversOther (5th and greater), LargeStreams (Strahler order 3rd, 4th), and SmallStreams
(Strahler order 1st, 2nd). Three ecological reporting regions (UMW, NPL, SPL) involve a smaller
number of states and were allocated fewer sites than the other six regions (NAP, SAP, CPL, TPL,
WMT, XER). Given these expected sample sizes, the number of sites for each state was allocated
proportional to the four medium resolution NHD Plus perennial river and stream category lengths in
each state for each ecological reporting region. Adjustments to the number of sites for states were
made to ensure that each state had a minimum of 20 sites from the NRSA09 and NRSA14 designs.
The final number of sites was also adjusted to ensure that a total of 1,808 unique sites were selected.
3.4.3 Stratification
The survey design is explicitly stratified by state for both the NRSA09 and NRSA14 designs.
3.4.4 Multi-Density Categories
Within each state, unequal probability of selection was based on river and stream categories as well as
ecological reporting regions.
24
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 3.2 Expected sample size for NRSA14 Design
Ecological
Region
Expected Sample Size NRSA14
RiversMajor
RiversOther
LargeStreams
SmallStreams
Total
NAP
29
30
33
35
127
SAP
29
30
33
35
127
CPL
29
30
33
35
127
TPL
29
30
33
35
127
XER
29
30
33
35
127
WMT
29
30
33
35
127
UMW
18
19
19
22
78
NPL
18
19
19
22
78
SPL
18
19
19
22
78
Total
228
237
255
276
996
3.4.4.1 NRSA09 Design
The target and sampled sites from NRSA 2008-09 were placed in sitelD order within a state and
within the River and Stream categories. The sites required for these categories were then selected as
the first set of sites within that list required to meet the sample size requirements. That is, the sites
were selected with equal probability within the categories.
The original NRSA 2008-09 survey design used unequal probability categories defined separately
for streams (1st to 4th order) and rivers (5th to 10th order). For the stream category, within each state
unequal selection probabilities were defined for 1st, 2nd, 3rd, and 4th order streams so that an equal
number of sites would occur for each order. Then these unequal selection probabilities were
adjusted by the nine ecological reporting regions so that an equal number of sites would occur in
each region. For the river category, unequal selection probabilities were defined for 5th, 6th, 7th, and
8th+ order rivers. Then these unequal selection probabilities were adjusted by the nine ecological
reporting regions so that an equal number of sites would occur in each region.
3.4.4.2 NRSA 14 design
The unequal probability of selection categories were the combination of state, ecological reporting
region, and the four river and stream categories (RiversMajor, RiversOther, LargeStreams, and
SmallStreams).
3.4.5 Oversample and Site Replacement
Both the NRSA09 and NRSA14 designs include a set of oversample sites to be used when a base site
cannot be sampled for any reason. The two designs have six categories within each state that are the
basis for the oversample and site replacement:
25
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 3.3 Oversample replacement categories
Replacement Category
Base Sites
Replace by Oversample
Sites
NRSA09 Rivers
Base_09River &
Base09River RVT
Over 09River
NRSA09 Streams
Base_09 Stream &
B ase09 StreamRVT
Over_09 Stream
NRSA14 Rivers Major
Base_NewRiversMaj or
Over_NewRiversMaj or
NRSA14 Rivers Other
B ase_N ewRi vers Other
Over_NewRiversOther
NRSA14 Large Streams
Base_NewLargeStreams
Over_N ewLarge Stream s
NRSA14 Small Streams
Base_NewSmallStreams
Over_NewSmall Streams
Sites within each state and above six categories are provided in sitelD order and the replacement
must be in sitelD order. The Base09River_RVT and Base09Stream_RVT sites identify the four sites
within each state that must be visited twice in 2013-14. If one of those sites cannot be sampled, then
the next site within the category then becomes a site to be visited twice.
3.4.6 State Designs
States may elect to implement a state-wide survey design in collaboration with NRSA. The above
survey design describes the national survey design and sets the required number of sites that must be
sampled within each state and the six design categories. There are two general types of state scale
surveys. The first type is one where a state may simply sample additional sites from the over NRSA
sample list of sites within their state to achieve a minimum of 50 sites. The second type is where the
state has state-specific survey design requirements. In this case a new survey design for the state is
completed that meets both the national and state survey design requirements. The new design will
include the NRSA09 design resample sites. This new survey design replaces the current national
design sites for the state, and sites that are not part of the NRSA09 set of resample sites can be
sampled in whatever manner that supports the state's monitoring program.
3.5 Evaluation Process
The survey design weights in the design file assumed that the survey design was implemented as
designed. To achieve the planned sample size, we replaced sites that could not be sampled with
oversamples as described above. Because some sites were replaced, the original survey design
weights are no longer correct and EPA statisticians had to adjust the weights. This weight
adjustment process required the statisticians knowing what happened to each site in the base
design and the oversample sites (e.g., was the site sampled or dropped and if dropped why?).
EvalStatus (evaluation status) was initially set to "NotEval" to indicate that the site had yet to be
evaluated for sampling. When a site was evaluated for sampling, then the EvalStatus for the site
26
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
was changed. Recommended codes are provided in Table 3.4.
Table 3.4 Recommended Codes for Evaluating Sites
EvalStatus
Code
Name
Meaning
TS
Target Sampled
Site was a member of the target population and was sampled
LD
Landowner Denial
Landowner denied access to the site
PB
Physical Barrier
Physical barrier prevented access to the site
NT
Non-Target
Site was not a member of the target population
NN
Not Needed
Site was a member of the oversample and was not evaluated for
sampling
Other codes
Other codes were often useful. For example, rather than use
NT, the status may include specific codes indicating why the
site was non-target.
3.6 Implementation of the design
For NRSA 2013-14, 4,566 design sites were evaluated. Of these 1,853 were evaluated as target and
sampled, with 192 sites sampled twice. The remaining sites were dropped and replaced for various
reasons (Table 3.5). The margin of error for national estimates was +/- 3% and for ecoregion
estimates was +/- 15% with 95% confidence. For the difference analysis, estimates had a margin of
error of + /- 5% at the national level and +/- 18% at the ecoregional level with 95% confidence. A
minimum of 20 sites were sampled in each state.
Table 3.5 Evaluation Status of Dropped Sites
Category
Number of sites dropped
Canal
285
Impounded
89
Inaccessible
385
Landowner_NoAccess
794
MapError
67
NonPerennial
755
NonTarget_Other
10
Target_Not_Sampled
149
Tidal
97
Wetland
82
3.7 Statistical Analysis
Any statistical analysis of the data must incorporate information about the monitoring survey
design. For NRSA, when estimates of characteristics for the entire target population are
computed, the statistical analysis must account for the stratifications and unequal probability
selection in the design. Procedures for doing this are available from the Aquatic Resource
Monitoring Web page (https: / /archive.epa.gov/nheerl/arm/web /html/index.html). A
27
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
statistical analysis library of functions to do common population estimates in the statistical
software environment R is available from the webpage. In the NRSA 2013-14 Site
Information data file, the adjusted weights used to calculate national condition estimates are
in the column "WGT_EXT_SP"
3.8 Literature Cited
Benke, A.C., and C.E. Cushing. 2005. Rivers of North America. Elsevier Academic Press, Burlington,
MA. 1168 pp
Diaz-Ramos, S., D. L. Stevens, Jr, and A.R. Olsen. 1996. EMAP Statistical Methods Manual.
EPA/620/R-96/002, U.S. Environmental Protection Agency, Office of Research and
Development, NHEERL-Western Ecology Division, Corvallis, Oregon.
Horn, C.R., and W.M. Grayman. 1993 Water-quality modeling with EPA reach file system. Journal of
Water Resources Planning and Management, 119, 262-74.
Stevens, D.L., Jr. 1997. Variable density grid-based sampling designs for continuous spatial
populations. Eniironmetrics, 8:167-95.
Stevens, D.L., Jr., and A.R. Olsen. 1999. Spatially restricted surveys over time for aquatic resources.
Journal of Agricultural, Biological, and Environmental Statistics, 4:415-428
Stevens, D.L., Jr., and A.R. Olsen. 2003. Variance estimation for spatially balanced samples of
environmental resources. Environmetrics 14:593-610.
Stevens, D.L., Jr., and A.R. Olsen. 2004. Spatially-balanced sampling of natural resources in the
presence of frame imperfections. Journal of American Statistical Association-. 99:262-278.
Strahler, A.N. 1957. Quantitative Analysis of Watershed Geomorphology. Trans. Am. Geophys. Un. 38,
913-920.
28
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
4 Selection of Sites to Establish Reference Conditions
The selection of least-disturbed reference sites described here for macroinvertebrate assessment (and
the initial screen for fish) was used with modifications for determining reference sites for fish, water
chemistry, and physical habitat analysis. See Section 5, Section 6, Section 7, and Section 8 for additional
details on the reference site process used for these indicators.
One way to assess current quality it to compare data to a benchmark. For a number of indicators, the
NARS assessments apply a reference approach, in which the least-disturbed reference sites in each
region of the U.S. are used to establish benchmarks for assessing quality at other sites. The Least-
disturbed condition approach attempts to capture the best available chemical, physical and biological
habitat conditions given the current state of the landscape. Data from reference sites were used to
select metrics for benthic macroinvertebrate and fish multimetric indices (MMI), develop benthic
macroinvertebrate Observed to Expected ratio (O/E) models, and define the ecoregion-specific
benchmarks use in the NARS analyses. This chapter describes the methodology used to select the
reference sites including the sources of potential reference sites; the chemical and physical screens; and
geospatial screens for assessing the quality of the benthic macroinvertebrate assemblage.
Applying the reference approach involves selecting and identifying least-disturbed reference sites to
establish a reference distribution from which benchmarks can be set. Using this approach, NRSA used
the 5th/25th or the 75th/95th percentiles of the reference distribution for setting good/fair/poor
benchmarks for a number of indicators (Hughes et al. 1986; USEPA 1996). See Section 5, Section 6,
and Section 7 for additional details.
In the first nationwide NARS assessment, the 2004 Wadeable Streams Assessment (WSA; USEPA,
2006), the primary biological indicator was stream macroinvertebrate assemblages. Reference sites were
compiled by filtering WSA sample sites for disturbance using a series of abiotic variables. Additional
reference sites were needed and were obtained from other state, university, and federal monitoring
programs. This pool of potential reference sites was then assessed for uniformity in site quality and
comparability of macroinvertebrate sample data. Ultimately, 1,625 sites were used to set reference
benchmarks for the WSA (Herlihy et al., 2008). These reference sites were used to develop a
macroinvertebrate multimetric index or MMI (Stoddard et al., 2008) and an observed/expected (O/E)
index generated from predictive models (Yuan et al., 2008) as assessment measures for WSA.
The NRSA 2008-09 analysis used the reference sites from WSA and we obtained new reference site
data from additional hand-picked and probability sites sampled during NRSA 2008-09. The NRSA
MMI and O/E approaches were revised to take advantage of these additional reference sites that
included both river and stream sites. For NRSA 2013-14 analysis, additional reference sites were
identified by filtering the 2013-14 sample for disturbance using the same WSA process. These new
reference sites were utilized in updating the reference-based benchmarks for the fish MMI and three of
the four physical habitat indicators. One of our goals is to stabilize the benchmarks we use across the
29
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
survey years. As part of that process, we wanted to analyze whether inclusion of a larger pool of
reference sites would indicate that a change might be needed or whether we could retain the
benchmarks developed and implemented in NRSA 2008-09. A comparison of existing NRSA 2008-09
benchmarks was made against new benchmarks calculated by adding new NRSA 2013-14 reference
sites. Based on this comparison, we decided the difference was small enough that we did not need to
update the benchmarks and that we could proceed with stabilization of the benchmarks for these
indicators.
4.1 Sources of Reference Sites
The fish and macroinvertebrate reference sites used in the NRSA came from four major activities:
1. First, we used sites sampled during the NRSA using consistent sampling protocols and analytical
methods. These included both sites selected from the probability sample and sites hand-picked
by best professional judgment and sampled using NRSA methods as part of the NRSA.
Analysts applied a three-tiered, pre-screening approach to select hand-picked sites as potential
reference sites for the NRSA.
First, sites throughout the country that were submitted as least disturbed by states,
academics, USG S, and EPA Regions were screened using a quantitative disturbance
score for the local watershed (the area draining to the reach segment).
Sites were then sent to the EPA Landscape Ecology Lab for a quantitative disturbance
score for the cumulative watershed (includes the reach and all upstream reaches).
Finally, the top 300 potential reference sites were ranked using a standardized qualitative
visual assessment of disturbance using Google Earth or ArcGIS at the 1: 24,000 and 1:
3,000 scales.
In the end, we sampled approximately 200 of these hand-selected river and stream sites
that covered the nine ecoregions and ranked high across all screens.
2. In addition to the sites sampled in the NRSA, we obtained data for potential reference site
from USGS' National Water-Quality Assessment Program (NAWQA), EPA Region 7, the
State of Wisconsin, and the State of Oklahoma. These data included fish and
macroinvertebrate assemblage data as well as physical and chemical habitat data.
3. Benthic macroinvertebrate reference site data also came from the 1,655 wadeable stream sites in
the EPA WSA. In the WSA, reference sites were obtained from two different approaches: first
by screening the WSA survey data for physical and chemical criteria in the same manner
described in #1 above, and second from macroinvertebrate data provided by other agencies,
universities, or states from sites that were deemed to be suitable as reference sites by best
professional judgment. These sites either were sampled with the same methodology as the WSA
or had field and lab protocols with enough similarities that the data analysis group determined
30
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
that the data were comparable. The reference sites from this second approach were only used in
developing an MMI for benthic macroinvertebrate samples, not for setting any benchmarks.
The WSA reference site screening process and data sources are described in detail in Herlihy et
al. (2008). In Table 4.1, the first two data columns summarize the number of available WSA
benthic macroinvertebrate reference sites by ecoregion.
4. We also pulled in additional fish reference site data from stream and river sites used by Herlihy et
al. (2006) in a national analysis of fish assemblage data. The screening process used to define
reference sites is described in Herlihy et al (2006) and defined in detail in Appendix 1 of that
document. The Herlihy etal. (2006) study only used the first two years of data from EMAP
(Environmental Monitoring and Assessment Program)-West. For NRSA, reference fish data
from the last three years of EMAP-West were also available and were included as well. Final
numbers of reference sites and screening used to refine the fish reference population are
outlined in Chapter 6.
Table 4.1 Macroinvertebrate reference sites available for use in the NRSA
\\ S.\ Ac(i\ ilics
NRSA Acti\iIies
\\ S A
\\ S A
NRSA
N RSA
l-'.co region
l'l\lcrn;il
Screened
I'lMcriiiil
Screened
1 ol;il
Northern Appalachians (NAP)
114
27
2
37
180
Southern Appalachians (SAP)
370
35
22
38
465
Coastal Plain (CPL)
112
15
3
46
176
Upper Midwest (UMW)
68
12
38
30
148
Temperate Plains (TPL)
124
38
50
22
234
Northern Plains (NPL)
10
18
3
47
78
Southern Plains (SPL)
56
21
51
34
162
Western Mountains (WMT)
335
129
4
40
508
Xeric Region (XER)
132
39
2
33
206
Total
1,321
334
175
327
2,157
4.2 Chemical and Physical Screens
To select reference sites from the all of the sites compiled as described in Section 4.1, we first used
chemical and physical data collected at each site {e.g., nutrients, turbidity, acidity, riparian condition)
to determine whether any given site is in least disturbed condition for its ecoregion. In the NRSA,
eight physical and chemical parameters were used to screen for reference sites, total N, total P,
chloride, sulfate, acid neutralizing capacity, turbidity, % fine substrate, and riparian disturbance
index. If a site exceeded the screening value for any one stressor it was dropped from reference
consideration.
Given that expectations of least-disturbed condition vary across ecoregions, the criteria values for
31
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
exclusion varied by ecoregion. The nine aggregate level III ecoregions developed for the WSA
assessment were used to regionalize reference conditions. Ecoregional specific screening criteria in
the Western Mountains ecoregion was broken into three finer-scale ecoregion subgroups for
screening to match EMAP-West's use of a somewhat finer spatial scale.
As noted in Section 4.1, in addition to the sites sampled in the NRSA, we obtained possible
reference site external data from four other agencies. Data from these external surveys were
screened for physical and chemical criteria using the same criteria used for NRSA sample sites in
Table 4.2 using whatever screening data were available in each survey.
All sites in the NRSA (both probability and hand-picked, boatable and wadeable) and the added
external data that passed all criteria were considered to be candidate reference sites for the NRSA
assessment. The number of reference sites that passed this screening is summarized in Table 4.1.
These reference sites include both fish and macroinvertebrate data. The NRSA did not use data on
the biological assemblages themselves for any screening as these are the primary components of the
stream and river ecosystems being evaluated, and to use them would constitute circular reasoning.
Note that the Rapid Bioassessment Protocol (RBP) physical habitat score was used as a filter in
WSA but was not available in the NRSA data to use as a screen. The six ecoregions in the top halfof
the table were used in WSA and reported in Herlihy et al. (2008), the ecoregions in the bottom half
of the table were screened using criteria developed in EMAP-West.
Sites were also screened using the criteria in Table 4.3 to identify sites that could be used to test
responsiveness in method and indicator development.
4.3 Geospatial Screens
As a final screen, all sites that passed the chemical and physical screens were then screened using
three additional landscape-GIS screening criteria. These screens included a dam influence index,
urbanization influence, and agricultural influence.
The dam influence index (DII) was used to assess the influence of upstream dams and the largest
reservoir on the current list of potential reference sites. The complete watershed was assessed for
any of the sites with a watershed boundary with a maximum distance of less than 200 km upstream
of the sampling point. Any site that had a watershed with a distance greater than 200 km upstream
of the sample point, had a wedge-shaped area assessed until 200 km upstream was reached. A cut-
off distance of 200 km upstream was used because it is unlikely land use activities occurring greater
than 200 km upstream will directly influence a given sample reach downstream. For example, a
sample reach on the lower Mississippi is more likely to be influenced by more proximal land use
activities than by land use activities in Montana, even though the Missouri River occuring within
Montana is part of the upstream watershed of the lower Mississippi. For all watersheds and wedges
assessed, a calculation of the volume of the largest reservoir, the number of dams, and an index
32
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
that weighted the maximum reservoir volume within the watershed or wedge by its proximity to
the sample point was conducted. Bach upstream reservoir was inversely weighted by its upstream
flow distance from the sample point as:
, D flow ^
Wj = e De folding'
where Dflow is the flow distance to the sample site, and Defoia&gis an e-folding value that determines
the rate at which the weight exponentially decreases (here 100 km). 1)11 equals the largest distance-
weighted volume within the watershed:
DII= max(Wj * Dj)
where Dt = reservoir volume (km3). The criteria for dropping a potential reference site was a DII
value equal to or greater than one.
Percent urbanization and agricultural influence were assessed within a 1 km2 area around the mid-
point of the sampled stream segment. To conduct this analysis a 1 km2 radius buffer around the
mid-point was overlaid onto the National Land Cover Database 2006 (USGS 2011) to calculate the
percentage of urban land cover and percent row crop, as defined by the NLCD (Figure 4.1). The
criteria for dropping a potential reference sites was any greater than 5% urban land cover and 15%
agricultural (row crop) land cover. The land cover percentages used for consistent screening of
near-reach human influence were based on best professional judgement.
A B
Figure 4.1 Examples of percent urban (A, 60%) and row crop (B, 72%) from NLCD
4.4 Literature Cited
Herlihy, A.T., R.M. Hughes, and J.C. Sifneos. 2006. National clusters of fish species assemblages m
the conterminous. United States and their relationship to existing landscape classification
schemes, pp. 87-112. In: Hughes, R.M., L. Wang, and P.W. Seelbach (Eds ). Influences of
Landscapes on Stream Habitats and Biological Assemblages. AmericanFisheries Society
Symposium 48, Bethesda, Maryland.
33
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Herlihy, A.T., S.G. Paulsen, J. Van Sickle, J.L. Stoddard, C.P. Hawkins, and L.L. Yuan. 2008.
Striving for consistency in a national assessment: the challenges of applying a reference
condition approach at a continental scale. Journal of the North American Benthological Society
27:860-877.
Hughes, R.M., D.P. Larsen, and J.M. Omernik. 1986. Regional reference sites: A method for
assessing stream potentials. Environmental Management. 10:629-635.
USEPA. 1996. Biological Criteria Technical Guidance for Streams and Small Rivers. EPA 822-B-
96-001. US Environmental Protection Agency, Office of Water, Office of Science and
Technology, Washington, DC.
USEPA. 2006. Wadeable Streams Assessment: a collaborative survey of the Nation's streams. EPA/641/B-
06/002, U.S. Environmental Protection Agency, Washington, D.C.
USGS. 2011. National Land Cover Database (NLCD) 2006 Land Cover Conterminous United States:
U.S. Geological Survey data release.
Yuan, L.L, C.P. Hawkins, and J. Van Sickle. 2008. Effects of regionalization decisions on an O/E
index for the US national assessment. Journal of the North American Benthological Society 21:892-
905.
34
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 4.2 Criteria for eight chemical and physical habitat filters used to identify the candidate
least disturbed reference sites for benthic macroinvertebrate and fish indicators for each of the
nine aggregate ecoregions for NRSA. A site must pass all eight filters to be considered a least-
disturbed reference site.
Filter criterion
NAP
SAP
CPL
UMW
TPL
SPL
NPL
XER
WMT-SWe
WMT-
SRocke
WMT-
Nrock/Pacifice
Total P (|ig/L)
<20
<20
<75
<50
<100
<150
<150
<50
<50
<25
<25
Total N (fj.g/L)
<750
<750
>2500
<1000
<3000
<4500
<4500
<1500
<750
<750
<750
CP (|ieq/L)
<250a
<200
-
<300
<2000
<1000
<1000
<1000
<300
<200
<200a
SO42- (|ieq/L)
<250
<400
<600
<400
-
-
-
-
-
<200
<200
ANC (|ieq/L) +
DOC (mg/L)b
>50 + >5
>50 + >5
>50 + >5
>50 + >5
>50 + >5
>50 + >5
>50 + >5
>50+>5
>50 + >5
>50 + >5
>50 + >5
Turbidity (NTU)
<5
<5
<10
<5
<50
<50
<50
<25
<5
<5
<5
Riparian Disturbance
Indexc
<2
<2
<2
<2
<2
<2
<2
<1.5
<0.5/<1.5d
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 4.3 Criteria for eight chemical and physical habitat filters used to identify the candidate
most-disturbedฎ sites for each of the nine aggregate ecoregions for NRSA. A site needed to
pass one of the eight filters to be considered a most-disturbed site.
Filter criterion
NAP
SAP
CPL
UMW
TPL
SPL
NPL
XER
WMT-SWe
WMT-
SRocke
WMT-
Nrock/Pacifice
Total P (ng/L)
>100
>100
>250
>150
>500
>500
>500
>150
>150
>100
>100
Total N (ng/L)
>3500
>3500
>8000
>5000
>15000
>10000
>10000
>5000
>1500
>1500
>1500
CI" (neq/L)
>10000
>1000
-
>2000
>5000
>5000
>5000
>5000
>1000
>1000
>1000
S042" (neq/L)
>1000
>1000
>4000
>2000
-
-
-
-
-
>1000
>1000
ANC (neq/L) +
DOC (mg/L)b
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
<0 +<5
Turbidity (NTU)
>10
>20
>50
>30
>100
>100
>100
>75
>10
>10
>10
Riparian Disturbance
Index0
>4
>4
>4
>4
>4
>3
>3
>3
>3
>3
>3
% fine substrate
>75
>75
>95
>90
>100
>99
>99
>90
>50
>50
>50
aA set of most-disturbed sites in each ecoregion are needed to test metric and MMI responsiveness
compared to least-disturbed. The criteria in Table 4.3 are the screening factors used to identify a set
of most-disturbed sites in each ecoregion.
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
5 Benthic Macroinvertebrates
Benthic macroinvertebrates were collected using a D-frame net with 500 [j.m mesh openings at 11
transects equally distributed along the targeted reach. Samples were composited from the 11
transects and the material was field preserved with -95% ethanol. Organisms were enumerated and
identified to the lowest possible taxonomic level (generally genus) using specified standard keys and
references (see the NRSA 2013-14 Field Operations Manual and Laboratory Operations Manual for
additional details). Benthic macroinvertebrate counts, metrics, and multimetric index condition from
NRSA are available to download from the NARS datawebpage - https://www.epa.gov/national-
aquatic-resource-surveys/data-national-aquatic-resource-surveys
The taxonomic composition and relative abundance of different taxa that make up the benthic
macroinvertebrate assemblage present in a stream have been used extensively in North America,
Europe, and Australia to assess how human activities affect ecological condition (Barbour et al. 1995,
1999; Karr and Chu 1999). As explained in general terms in the NRSA 2008-09 Technical Report
(USEPA 2016; see Section 5.2) two principal types of ecological assessment tools to assess condition
based on benthic macroinvertebrates are currently prevalent: multimetric indices and predictive
models of taxa richness. The purpose of these indicators is to present the complex community
taxonomic data represented within an assemblage in a way that is understandable and informative to
resource managers and the public. The following sections provide an overview of the approaches
used to develop ecological indicators based on benthic macroinvertebrate assemblages, followed by
details regarding data preparation and the process used for each approach to arrive at a final
indicator. The same analyses and benchmarks were used in NRSA 2008-09 and NRSA 2013-14.
5.1 Overview
Multimetric indicators have been used in the U.S. to assess stream condition based on fish and
macroinvertebrate assemblage data {e.g., Karr and Chu, 1999; Barbour et al., 1999; Barbour et al.,
1995). The multimetric approach involves summarizing various assemblage attributes {e.g.,
composition, tolerance to disturbance, trophic and habitat preferences) as individual "metrics" or
measures of the biological community. Candidate metrics are then evaluated for various aspects of
performance and a subset of the best performing metrics are then combined into an index, referred
to as a multimetric index or MMI. For NRSA 2013-14 and NRSA 2008-09, the benthic
macroinvertebrate MMI developed in the WSA was used to generate the population estimates used
in the assessment. The WSA MMI is detailed in Stoddard et al. (2008).
The predictive model approach was initially developed in Europe and Australia, and is becoming
more prevalent within the U.S. The approach estimates the expected taxonomic composition of an
assemblage in the absence of human stressors (Hawkins et al., 2000; Wright, 2000), using a set of
"least-disturbed" sites and other variables related natural gradients (such as elevation, stream size,
stream gradient, latitude, longitude). The resulting models are then used to estimate the expected
taxa composition (expressed as taxa richness) at each stream site sampled. The number of
37
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
expected taxa observed at a site is compared to the total number of expected taxa as an
observed:expected ratio (O/E index). Departures from a ratio of 1.0 indicate that the taxonomic
composition in a stream sample differs from that expected under less disturbed conditions.
5.2 Data Preparation
5.2.1 Standardizing Counts
The number of individuals in a sample was standardized to a constant number to provide an
adequate number of individuals that was the same for the most samples and that could be used for
both multimetric index development and O/E predictive modeling index. A subsampling technique
involving random sampling without replacement was used to extract a true "fixed count" of 300
individuals from the total number of individuals enumerated for a sample (target lab count was 500
individuals). Samples that did not contain at least 300 individuals were used in the assessment
because low counts can indicate a response to one or more stressors. Only those sites with at least
250 individuals, however, were used as least-disturbed reference sites.
5.2.2 Operational Taxonomic Units
For the predictive model approach, it was necessary to combine taxa to a coarser level of common
taxonomy. This new combination of taxa is termed an "operational taxonomic unit" or OTU, and
results in fewer taxa than are present in the initial benthic macroinvertebrate count data.
5.2.3 Autecological Characteristics
Autecological characteristics refer to specific ecological requirements or preferences of a taxon for
habitat preference, feeding behavior, and tolerance to human disturbance. These characteristics are
prerequisites for identifying and calculating many metrics. A number of state/regional organizations
and research centers have developed autecological characteristics for benthic macroinvertebrates in
their region. For the WSA and NRSA, a consistent "national" list of characteristics that consolidated
and reconciled any discrepancies among the regional lists was needed before certain biological
metrics could be developed and calibrated and an MMI could be constructed. The same
autecological information used in WSA was used in NRSA 2008-09 and 2013-14.
Members of the data analysis group pulled together autecological information from five existing
sources: (1) the EPA Rapid Bioassessment Protocols document; (2) the USGS National Ambient
Water Quality Assessment (NAWQA) national and northwest lists; (3) the Utah State University list;
(4) the EMAP Mid-Atlantic Highlands (MAHA); and (5) the EMAP Mid-Atlantic Integrated
Assessment (MAIA) list. These five were chosen because they were thought to be the most
independent of each other and the most inclusive. A single national-level list was developed based on
the decision rules described in the following sections.
5.2.3.1 Tolerance Values
38
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Tolerance value assignments followed the convention for macroinvertebrates, ranging between 0
(least tolerant or most sensitive) and 10 (most tolerant). Foreach taxon, tolerance values from all
five sources were reviewed and a final assignment made according to the following rules:
If values from different lists were all <3 (sensitive), final value = mean.
If values from different lists were all >3 and <7 (facultative), final value = mean.
If values from different lists were all >7 (tolerant), final value = mean.
If values from different lists spanned sensitive, facultative, and tolerant categories,
best professional judgment was used, along with alternative sources ofinformation
(if available) to assign a final tolerance value.
Tolerance values of 0 to <3 were considered "sensitive" or "intolerant." Tolerance
values >7 to 10 were considered "tolerant," and values in between were considered
"facultative."
5.2.3.2 Functional Feeding Group and Habitat Preferences
In many cases, there was agreement among the five data sources identified in Section 5.2.3. When
discrepancies in functional feeding group (FFG) or habitat preference ("habit") assignments among
the five primary data sources were identified, a final assignment was made based on the most
prevalent assignment. In cases where there was no prevalent assignment, the workgroup examined
why disagreements existed, flagged the taxon, and used best professional judgment to make the final
assignment.
5.3 Multimetric Index Development
5.3.1 Regional Multimetric Development
The same autecology and taxonomic resolution used in WSA was applied to the NRSA
macroinvertebrate 300 fixed count data to calculate the community metrics used to calculate the
MMI. In the WSA, a best ecoregional MMI was developed by summing the six metrics that
performed best in that ecoregion (the national aggregate nine ecoregions). Each of the six metrics
was scored on a 010 scale by interpolating metrics between a floor and ceiling value. The six metric
0-10 point scaled scores were then summed and normalized to a 0100 scale by multiplying by
100/60 to calculate the final MMI. Details of this process are described in Stoddard et al. (2008).
The final metrics used in each ecoregion, metric direction, and floor and ceiling values are
summarized in Table 5.1. Scoring equations are different depending on if the metric responds
positively (high values good) or negatively (high values bad) with disturbance. For positive metrics,
values above the ceiling get 10 points, and values below the floor get 0 points. For negative metrics,
values above the ceiling get 0 points, and values below the floor get 10 points. The interpolation
equations for scoring the 0-10 points for metrics between the floor and ceiling values are:
Positive Metrics: Metric Points = 10 * ((metric value-floor)/(ceiling-floor))
39
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Negative Metrics: Metric Points = 10 * (1 - ((metric value-floor)/(ceiling-floor))).
The MMI used in the NRSA report is identical to the WSA MMI in terms of metrics and scoring.
Based on NRSA revisit data, the MMI had a S:N ratio of 2.8 and a pooled standard deviation of 10.0
(out of 0100).
40
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 5.1 Six benthic community metrics, scoring direction, and floor and ceiling values used in
calculating the NRSA and WSA MMI in each of the nine aggregate ecoregions.
Ecoregion
Direction
Metric
Floor
Ceiling
Negative
Non-Insect % Individuals
0.70
73.0
Positive
Shannon Diversity
1.62
3.31
CPL
Positive
Shredder Taxa Richness
1
9
Positive
dinger % Taxa Richness
14.3
54.8
Positive
EPT Taxa Richness
1
17
Negative
Tolerant % Taxa Richness
5.56
50.0
Positive
EPT % Taxa Richness
9.52
57.6
Negative
% Individuals in Top 5 Taxa
37.2
76.2
NAP
Positive
Scraper Taxa Richness
3
12
Positive
dinger % Taxa Richness
28.6
70.0
Positive
EPT Taxa Richness
3
24
Positive
PTV 0-5.9 % Taxa Richness
46.2
86.1
Positive
EPT % Taxa Richness
3.85
50.0
Positive
Shannon Diversity
1.10
3.07
NPL
Positive
Scraper Taxa Richness
1
6
Negative
Burrower % Taxa Richness
6.45
35.3
Positive
Ephemeroptera Taxa Richness
0
7
Positive
PTV 0-5.9 Taxa Richness
4
28
Positive
Ephemeroptera % Taxa Richness
5.41
28.6
Positive
Shannon Diversity
2.05
3.44
SAP
Positive
Scraper Taxa Richness
3
12
Negative
Burro wer % Taxa Richness
3.45
25.0
Positive
EPT Taxa Richness
5
25
Negative
Tolerant % Taxa Richness
2.44
27.6
Positive
EPT % Individuals
0.67
66.0
Positive
Shannon Diversity
1.16
3.27
SPL
Positive
Scraper Taxa Richness
1
8
Negative
Burro wer % Taxa Richness
5.0
36.1
Positive
EPT Taxa Richness
1
16
Positive
Intolerant Taxa Richness
1
8
Positive
EPT % Individuals
0.67
80.3
Positive
Shannon Diversity
1.41
3.17
TPL
Positive
Scraper Taxa Richness
1
9
Positive
dinger Taxa Richness
20
Positive
Ephemeroptera Taxa Richness
1
11
Negative
PTV 8-9.9 % Taxa Richness
4.35
33.3
Negative
Chironomid % Taxa Richness
11.2
50.8
Positive
Shannon Diversity
2.01
3.56
UMW
Positive
Shredder Taxa Richness
3
10
Negative
Burro wer % Taxa Richness
3.77
28.6
Positive
EPT Taxa Richness
4
22
Negative
PTV 8-9.9 %Taxa Richness
2.51
29.5
Positive
EPT % Taxa Richness
18.5
62.9
Negative
% Individuals in Top 5 Taxa
40.6
82.3
WMT
Positive
Scraper Taxa Richness
1
8
Positive
dinger % Taxa Richness
27.0
69.6
Positive
EPT Taxa Richness
6
23
Negative
Tolerant %Taxa Richness
2.27
25
Negative
Non-Insect % Individuals
3.33
36.0
Negative
% Individuals in Top 5 Taxa
44.7
92.3
XER
Positive
Scraper Taxa Richness
0
7
Positive
dinger % Taxa Richness
15.8
65.8
Positive
EPT Taxa Richness
1
18
Negative
Tolerant % Taxa Richness
3.57
36.4
41
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
5.3.2 Modeling of MMI Benchmarks
Previous large-scale assessments have converted MMI scores into classes of assemblage qualityby
comparing those scores to the distribution of scores observed at least-disturbed reference sites. If a
site's MMI score was less than the 5th percentile of the reference distribution, it was classified as
"poor" quality; scores between the 5th and 25th percentile were classified as "fair"; and scores in the
25th percentile or higher were classified as "good." This approach assumes that the distribution of
MMI scores at reference sites reflects an approximately equal, minimum level of human disturbance
across those sites. But this assumption did not appear to be valid for some of the nine WSA regions,
which was confirmed by state and regional parties at meetings to review the draft results.
For the WSA, the project team performed a principal components analysis (PCA) of the physical
habitat and water chemistry variables (Total P, Total N, pH, Chloride, Sulfate, Turbidity, %Fine
Substrate, Riparian Disturbance Index) that had originally been used to screen for biological
reference sites as described in Chapter 4. The first principal component (Factor 1) of this PCA well
represented a generalized gradient of human disturbance. MMI scores at the reference sites,
however, were weakly, but significantly, related to this disturbance gradient in some of the aggregate
ecoregions. Thus, MMI reference distributions from these regions may be biased downward,
because they include somewhat disturbed sites which may have lower MMI scores. As part of the
WSA, Herlihy et al. (2008) developed a process that used this PCA disturbance gradient to reduce
the effects of disturbance on benchmark values within the reference site population. The process
uses multiple regression modeling to develop adjusted benchmarks analogous to the 5th and 25th
percentiles of reference sites in each ecoregion based on the slope of the MMI-disturbance
relationship in each ecoregion.
These adjusted benchmarks were used in the WSA but were based on a fairly small sample size of
reference sites. To increase the sample size used in the regression model, the benchmark adjustment
process was rerun for NRSA using the original WSA reference sites plus the additional NRSA
reference sites identified in Section 4. As in the WSA analysis and other benchmark setting, we used
a 1.5*interquartile range (IQR) outlier screening test in each ecoregion to drop MMI outliers from
the analysis (sites with values outside the range of Q1-1.5*IQR or Q3+1.5*IQR were dropped).
This removed 6 sites from the analysis (all low; 3 in WMT, and 3 in XER). There were a grand total
of 647 least-disturbed reference sites used for the benchmark regression adjustment modeling and
the resulting regression statistics for each ecoregion are shown in Table 5.2. The process for
calculating these adjusted benchmarks and fitting the regression model is detailed in Herlihy et al.
(2008). Briefly, the process involves setting the goal for disturbance to the 25th percentile of the
Factor 1 disturbance score for reference sites in each ecoregion. The ecoregion MMI value at that
goal is predicted from the MMI-disturbance regression as:
MMIpred = (GOAL * SLOPE) + INTERCEPT
Then the percentiles to be used as the adjusted benchmarks are calculated assuming there is a
42
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
normal distribution around this predicted mean using the RMSE of the regression model as the
standard error,
Good-Fair 25th benchmark = MMIpred - 0.675 * RMSE
Fair-Poor 5th benchmark = MMIpred - 1.650 * RMSE
The resulting adjusted MMI benchmark values for the condition classes in each ecoregion used in
the NRSA report are given in Table 5.3.
Table 5.2 MMI-Disturbance Regression Model Statistics Used for Setting Benchmarks
Ecoregion
Number of
Reference
Sites
Factor 1
Goal*
Regression
RMSE
Regression
Slope
Regression
Intercept
CPL
32
-0.1501
14.55
0
64.74
NAP
56
-0.5247
14.55
-7.257
61.06
NPL
65
0.8723
14.55
-14.95
79.66
SAP
64
-0.5531
14.55
-7.257
50.78
SPL
43
0.7637
14.55
-7.257
50.84
TPL
49
1.045
14.55
-7.257
57.75
UMW
39
-0.1138
14.55
0
46.74
WMT
209
-1.326
14.55
-7.257
50.27
XER
90
-0.4628
14.55
-7.257
63.44
* The 25th percentile of Factor 1 score was the "goal" on the PCA factor 1 disturbance gradient for hindcasting
ecoregional benchmarks.
Table 5.3 Benchmark Values for the Nine Regional Benthic MMIs
Ecoregion
Good Benchmark
Poor Benchmark
CPL
>54.9
<40.7
NAP
>55.0
<40.9
NPL
>56.8
<42.6
SAP
>45.0
<30.8
SPL
>35.5
<21.3
TPL
>40.3
<26.2
UMW
>36.9
<22.7
WMT
>50.1
<35.9
XER
>57.0
<42.8
*Any site with an MMI score that was not "good" or "poor" was considered "fair."
5.4 Predicted O/E Modeling
In addition to the benthic macroinvertebrate MMI approach, predictive O/E modeling was used to
assess benthic macroinvertebrate condition. The O/E model compares the observed benthic
assemblage at a site to an expected assemblage derived from the reference sites.
Stressors and anthropogenic impacts typically lead to a reduction in the number of taxa that are
43
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
expected to be present under reference conditions. The predictive model approach is used by
several states and is a primary assessment tool of Great Britain and Australia. The O/E ratio
predicted by the model for any site expresses the number of taxa found at that site (O), as a
proportion of the number that would be expected (E) if the site was in least disturbed condition.
Ideally, a site in reference condition has O/E = 1.0. An O/E value of 0.70 indicates that 70% of
the "expected" taxa at a site were actually observed at the site. This is interpreted as a 30% loss of
taxa relative to the site's predicted reference condition. However, O/E values vary among
reference sites themselves, around the idealized value of 1.0, because such sites rarely conform to
an idealized reference condition, and because of model error and sampling variation. The standard
deviation of O/E (Table 5.4) indicates the breadth of O/E variation at reference sites. Thus, the
O/E value of an individual site should not be interpreted as (1 taxa loss) without taking account
of this variability in O/E. Individual O/E values are most reliably interpreted relative to the entire
O/E distribution for reference sites.
A nationally distributed collection of reference sites was first identified, drawn from a pool of sites
whose macroinvertebrates were sampled using NRSA protocols. This pool included only NRSA,
WSA, EMAP-West, STAR-Hawkins, USGS NAWQA, and MAHA/MAIA sites. One hundred
reference sites were set aside to validate the models, and the remaining reference sites were used to
calibrate the models (Table 5.4). Each site contributed a single sampled macroinvertebrate
assemblage to model calibration and validation. Each sampled macroinvertebrate assemblage
comprising more than 300 identified individuals was randomly subsampled to yield 300 individuals.
300-count subsamples were used to build models and assess all NRSA sites.
The predictive modeling approach assumes that expected assemblages vary across reference sites
throughout a region, due to natural (non-anthropogenic) environmental features such as geology,
soil type, elevation, and precipitation. To model these effects, the approach first classifies reference
sites based on similarities of their macroinvertebrate assemblages (Table 5.4). A random forest
model is then built to predict the membership of any site in these classes, using natural
environmental features as predictor variables (Table 5.4). The predicted occurrence probability of a
reference taxon at a site is then predicted to be the weighted average of that taxon's occurrence
frequencies in all reference site classes, using the site's predicted group membership probabilities in
the classes as weights. Finally, E for any site is the sum, over a subset of reference taxa, of predicted
taxon occurrence probabilities. O is the number of taxa in that subset that were observed to be
present at the site. The subset of reference taxa used for any site was defined as those taxa with
predicted occurrence probabilities exceeding 0.5 at that site.
Final predictive models performed better than corresponding null models (no adjustment for
natural-factor effects), as judged by their smaller standard deviation of O/E across calibration sites
(Table 5.4). Similar to the IBI, two scaled approaches were used to develop the O/E model. A
national model was initially developed to predict taxa loss at sites. Three models were developed
for NRSA usage, together covering the contiguous USA (Table 5.4). The regional models
performed better and were used in the NRSA to predict taxa loss at the sites.
44
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 5.4 Benthic Macroinvertebrate Predictive Models
Model Niiiiic
Regions covered
lliislcrii lli^hliiiids
NAP, SAP
Pliiins iind l.owhinds
CPL, UMW, TPL, NPL,
SPL
Wesl
WMT, XER
Number of calibration
sites
297
241
659
Number of validation
sites
31
21
48
Number of site classes
17
16
Random Forest predictor
variables
Predicted mean summer
stream temperature,
watershed area, watershed
mean minimum annual
temperature, predicted
mean annual stream
temperature, watershed
mean annual temperature,
watershed mean
minimum precipitation
Predicted mean annual
stream temperature,
watershed mean date of
last freeze, watershed
mean soil permeability,
watershed mean runoff,
watershed maximum
elevation
Watershed area,
watershed mean annual
temperature, watershed
mean precipitation
accumulation, predicted
mean annual stream
temperature, watershed
mean maximum
temperature, watershed
mean elevation
Standard deviation of O/E
at calibration sites:
-- Predictive model
-- Null model
0.18
0.22
0.23
0.26
0.18
0.25
5.5 Literature Cited
Barbour, M.T., J. Gerritsen, G.E. Griffith, R. Frydenborg, E. McCarron, J.S. White, and M.L.
Bastian. 1996. A framework for biological criteria for Florida streams using benthic
macroinvertebrates. Journal of the North American Benthological Society 15:185-211.
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid bioassessmentprotocolsfor
use in streams and wadeable rivers:periphyton, benthic macroinvertebrates, andfish. EPA 841/B-
99/002, Office of Water. US Environmental Protection Agency, Washington,DC.
Hawkins, C.P., R.H. Norris, J.N. Hogue, and J.W. Feminella. 2000. Development and
Evaluation of Predictive Models for Measuring the Biological Integrity of Streams.
E cological Applications 10(5): 1456-1477.
Herlihy, A.T., S.G. Paulsen, J. Van Sickle, J.L. Stoddard, C.P. Hawkins, and L.L. Yuan. 2008.
Striving for consistency in a national assessment: the challenges of applying a reference
condition approach at a continental scale. Journal of the North American Benthological Society
27:860-877.
Karr, J. R., and E.W. Chu. 2000. Sustaining living rivers. Hydrobiologia 422/423:1-14.
Stoddard, J.L., A.T. Herlihy, D.V. Peck, R.M. Hughes, T.R. Whittier, and E. Tarquinio. 2008. A
process for creating multi-metric indices for large scale aquatic surveys. Journal of the
North American Benthological Society 27:878-891.
USEPA. 2016. National Rivers and Streams Assessment 2008-2009. Technical Report EPA 841-
R-l 6-008. U.S. Environmental Protection Agency, Office of Water and Office of
Research and Development, Washington, DC.
45
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Wright, J.F., 2000. An introduction to RIVPACS. In: Wright, J.F., D.W. Sutcliffe, and M.T.
Furse (Eds.). Assessing the Biological Quality of Fresh Waters: RIVPACS and Other Techniques.
Freshwater Biological Association, Ambleside, UK, pp. 1-24.
Yuan, L.L, C.P. Hawkins, and J. Van Sickle. 2008. Effects of regionalization decisions on an
O/E index for the US national assessment. Journal of the North American Benthological Society
27:892-905.
46
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6 Fish Assemblage
6.1 Background
Fish assemblages in streams and rivers offer several unique advantages to assess ecological quality,
based on their mobility, longevity, trophic relationships, and socioeconomic importance (Barbour et
al. 1999, Roset et al. 2007). For fish assemblages, assessing ecological quality has generally been
based on developing and using multimetric indices (MMIs), which are derivations of the original
Index of Biotic Integrity (IBI) developed by Karr (Karr 1981). There are numerous examples of
MMIs developed for fish assemblages in smaller streams (e.g., McCormick et al. 2001, Hughes et al.
2004, Bramblett et al. 2005, Roset et al. 2007) as well as for larger rivers (Lyons et al. 2001, Emery et
al. 2003, Mebane et al. 2003, Pearson et al. 2011).
6.1.1 Multimetric Indicator for NRSA 2008-09
For the NRSA 2008-09, we developed fish MMIs using predictive models of metric response (e.g.,
Oberdorff et al. 2002, Tejerina-Garro et al. 2006, Pont et al. 2007, Pont et al. 2009). This approach
essentially provided an estimate of expected quality (in terms of metric values) at individual sites,
rather than using a set of regional least-disturbed reference sites to define expected values for a
particular metric. Several studies concluded that the combined approach resulted in MMIs that
performed better in terms of their ability to discern deviation from expected condition (Oberdorff
et al. 2002, Tejerina-Garro et al. 2006, Pont et al. 2007, Pont et al. 2009, Hawkins et al. 2010a). Details
regarding the development and performance of these fish MMIs are presented in the technical
support document for NRSA 2008-09 (U.S. EPA 2016).
6.1.2 Multimetric Indicator for NRSA 2013-14
The fish MMIs developed and used for the NRSA 2008-09 assessment performed adequately in
terms of their responsiveness to disturbance and repeatability (USEPA 2016). However, there were
several major constraints associated with these MMIs. The two major constraints were: 1) the large
number of least-disturbed reference sites required to construct the predictive models of metric
response, which limited our ability to develop MMIs for smaller regions; and 2) the difficulty in
transferring the model outputs, R scripts, etc. to potential users to apply to their own data. These
constraints, along with new approaches to constructing and evaluating MMIs that became available
since the original fish MMIs were developed, led us to investigate using a more traditional approach
to MMI development. This more traditional approach adjusted metrics for watershed area using
linear regression if the effect was large enough.
The NRSA 2008-09 fish data were used to develop and evaluate fish MMIs based on the traditional
approach. We compared the performance of the traditional fish MMIs to the original model-based
MMIs (see Appendix 7.A), and then applied the traditional MMI approach to both the NRSA
2008-09 and 2013-14 data for use in the 2013-14 survey report.
47
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.1.3 Regionalization
We developed three original model-based fish MMIs for NRSA 2008-09, one for each of the three
climatic regions (Eastern Highlands, Plains and Lowlands, and the West; Figure 6.1). We
developed separate traditional fish MMIs for each of the nine NARS reporting regions for NRSA
2013-14 (Figure 6.2).
6.2 Methods
6.2.1 Field methods
Collection methods for fish are described in the NRSA 2013-14 field operations manuals (USEPA
2013a, b). Collection methods used for NRSA 2013-14 were essentially unchanged from those used
for NRSA 2008-09 (USEPA 2009). These minor changes included sorting fish into 6-inch size
classes instead of recording only minimum and maximum length, and minor text changes to help
clarify sampling procedures or field forms. Three variants of the basic sampling protocol (using
electrofishing) were used depending on the width of the stream and if it was wadeable. For
wadeable streams less than 12.5 m wide, a reach length equal to 40 channel widths was sampled for
fish. For larger wadeable streams (> 12.5 m wide), a reach length of 500 m or 20 channel widths
was sampled (whichever was longer). For non-wadeable streams and rivers, at least a reach length
of 20 channel widths was sampled. At large wadeable and non-wadeable sites, sampling continued
past the established reach length until 500 individuals were collected (or a reach length equal to 40
channel widths was sampled).
For fish, 2,261 site visits were initially available. These included 2,045 visits to 1,853 probability
sites and 216 hand-picked sites (single visit) that were evaluated as potential least-disturbed
reference sites (see Section 4.1). There were 192 revisits to a subset of the 1,853 probability sites
(either within a single year or across the two years of sampling). Fish sampling was attempted at
2,059 sites (not including revisits). A sufficient sample (based on length of reach sampled for fish
and the number of individuals collected) was obtained at 1,847 sites, with no fish collected at 64
sites. Seining only was conducted at 36 sites, and conditions prevented a sufficient sample from
being collected at 167 sites. No fish data were obtained from 133 sites, due to the lack of required
permits (66 sites), equipment failure (8 sites), site conditions (44 sites), loss of data after collection
(9 sites), or other reasons (6 sites).
6.2.2 Counting, Taxonomy, andAutecology
Fish were tallied and identified in the field, then released alive unless used for fish tissue or
vouchers. Voucher specimens were collected if field identification could not be accomplished.
Voucher samples of all species collected were also prepared at 10% of sites for each field
taxonomist. Voucher samples were sent to an independent taxonomist to evaluate taxonomic
proficiency of each field taxonomist. All names submitted on field data forms were reviewed and
48
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
revised when necessary to create a listing of nationally consistent common and scientific names.
Where possible, taxonomic names (common and scientific) were based on Nelson et al. (2004) and
Page et al. (2013). The online database FishBase (http://www.fishbase.org) served as a secondary
source of taxonomic names. In rare cases, a journal article of a newly described species was used.
Collection maps for each taxon were prepared and compared to published maps in Page and Burr
(2011) or alternative web publications for a few rare endemic species. A total of 631 unique taxa
were identified, excluding unknowns, hybrids, and amphibians. Amphibians were not used in the
fish MMIs but were retained in the database for potential use by other users of NRSA data.
Each taxon was characterized for several different autecological traits, based on available sources of
published information (e.g., McCormick et al. 2001, Goldstein and Meador 2004, Whittier et al.
2007b, Frimpong and Angermeier 2009). Traits included habitat guilds (lotic habitat and
temperature), trophic guild, reproductive guild, migration strategy, and tolerance to human
disturbance. A file of all fish taxa and their associated autecological assignments is available on the
NRSA website.
Assignments of native status were based primarily on shapefiles of individual species distribution
from NatureServe ([http://www.natureserve.org). Alternative sources included the USGS
Nonindigenous Species database (http://nas.er.usgs.gov), FishBase, published maps in Page and
Burr (2011), and relevant state fish publications (if available).
Because fish collected at a site cannot always be confidently identified to species, there is a risk of
inflating the number of species actually collected. For each sample, we reviewed the list of taxa to
determine whether they were represented at more than one level of resolution. For example, if an
"Unknown Catostomus" was collected, and it was the only representative of the genus at the site, we
assigned it as a distinct taxon. If any other species of the genus were collected, then we considered
the unknown as not distinct. We used only the number of distinct taxa in the sample to calculate
any metrics based on species richness.
49
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
NRSA Climatic Regions
Eastern Highlands
Plains and Lowlands
West
0 195 390
1,560
ฆ Kilometers
Figure 6.1 Aggregated Omernik ecoregions used to develop model-based fish MMIs for
NRSA 2008-09. A separate fish MMI was developed for each of the three climatic regions.
50
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
UMW,
TPL
Figure 6.2 Aggregated Omernik ecoregions used to develop traditional fish MMIs for NRSA
2013-14. A separate fish MMI was developed for each of the nine aggregated ecoregions.
NA P=Northern Appalachians, SAP=Southern Appalachians, CPL= Coastal Plains,
TPL=Temperate Plains, l.JMW= Upper Midwest, SPL=Southern Plains, NPL=Northern
Plains, XER=Xeric West, WMT=Western Mountains.
6.3 Fish multimetric index development
We used a consistent process to develop a multimetric index for fish for each of the nine
aggregated ecoregions. We used the sites from the NRSA 2008-09 to develop and evaluate the fish
MMIs, then calculated fish MMI scores for the NRSA 2013-14 data. We evaluated each metric for
its responsiveness to disturbance, i.e., its ability to discern between least-disturbed and most-
disturbed sites (following Stoddard et al. 2008). We then selected metrics representing different
dimensions of assemblage structure or tunction to include in the fish MMI based on responsiveness
and lack of correlation with other metrics, following W hi trier et al. (2007b) and Stoddard et, al.
(2008).
51
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.3.1 Least-Disturbed Reference Sites for Fish
We modified the base list of least-disturbed reference sites (Section 4) determined for NRSA to
eliminate additional fish samples that might not be representative of least-disturbed conditions (i.e.,
excluded sites where < 25 fish were caught or had >50% non-native individuals) (Table 6.1). The
final set of least-disturbed reference sites are identified in the NRSA database (variable
RT_NRSA_FISH=R).
To validate the fish MMIs and their component metrics, we identified a random subset of least-
disturbed sites (validation sites) within each aggregated ecoregion and excluded them from fish
MMI development. We set aside 29 validation sites in the Eastern Highlands (NAP=16, SAP=13),
66 sites in the Plains and Lowlands (CPL=10, NPL=16, SPL=13, TPL=14, UMW=13), and 23
sites in the West region (WMT=13, XER=10). We expected the distribution of fish MMI scores
calculated for the validation sites would be similar to the distribution of fish MMI scores calculated
for the calibration sites that were used to develop the fish MMIs.
6.3.2 Candidate Metrics
We calculated 162 candidate metrics (Appendix 7.B) representing the following dimensions of fish
assemblage structure and function (following Stoddard et al. 2008):
Nonnative species (ALIEN) based on presence in 8-digit USGS Hydrologic Units
Taxonomic composition (COMP)
Species richness (RICH)
Habitat guild (HABIT)
Life history/migratory strategy (LIFE)
Reproductive guild (REPRO)
Trophic guild (TROPH)
Tolerance (TOLER) to anthropogenic disturbance
The codes (in uppercase) for each category are used in the NRSA database to identify metric
categories. For nearly all metrics, we derived three variants based on all taxa in the sample and for
only native taxa in the sample: one based on distinct taxa richness, one based on the percent of
individuals in the sample, and one based on the percent of distinct taxa in the sample (potentially
yielding 6 different variants). For some trophic metrics, additional variants were derived using only
taxa that were not considered tolerant to disturbance. We included only those tolerance metrics
based on sensitive and tolerant taxa, because the "intermediate tolerance" assignments included
taxa with unknown tolerance.
6.3.3 Adjustment of Metric Response for Watershed Area
We used the set of least-disturbed reference sites in each aggregated ecoregion to evaluate whether
metrics should be adjusted for stream size. Many studies have shown that some metrics (especially
52
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
those based on species richness) vary naturally with stream size (e.g., Fausch et al. 1984, Simon and
Lyons 1995, McCormick et al. 2001). We used watershed area (in km2) as our measure of stream
size and compared the metric response to watershed area (transformed using loglO) using linear
regression. We used an Revalue >0.10 (following the rationale of Hawkins et al. (2010a) and Vander
Laan and Hawkins (2014)) in deciding whether or not to use the model-adjusted responses for a
particular metric. For metrics requiring adjustment, we used the residual values from the regression
as the adjusted metric response (Stoddard et al. 2008).
Table 6.1 Criteria used to select least-disturbed sites for use in developing the regional
NRSA fish multimetric indices (MMIs) based on 2008-09 data.
Criteria
Start with the base set of NRSA least-disturbed reference sites
Keep sites with fish
samples
Drop sites where seining was the only sampling method
Drop sites with insufficient sampling
Wadeable: Reach length sampled was less than 20 channel widths and less than
500 individuals were collected
Large Wadeable: Reach length sampled was less than 500 m and less than 500
individuals were
collected
Boatable: Reach length sampled was less than 20 channel widths
sampled
Drop sites with sufficient sampling where less than 30 individuals were collected
Drop sites with sufficient sampling where nonnative individuals comprised >50% of total
number of individuals collected
Drop non
-wadeable sites hand-selected from the EMAP-Western Pilot Study that were
sampled for fish. These sites were sampled using a much larger reach length (100 channel
widths) than the reach length used for NRSA (40 channel widths).
Final Number of Least-Disturbed Reference Sites
Calibration Sites
Validation Sites
Total
Northern
NAP
43
16
59
Appalachians
Southern
SAP
72
13
85
Appalachians
Coastal Plains
CPL
27
10
37
Northern Plains
NPL
33
16
49
Southern Plains
SPL
34
13
47
Temperate
TPL
31
14
45
Plains
Upper Midwest
UMW
48
13
61
Western
WMT
77
13
90
Mountains
Xeric West
XER
30
10
40
Total
395
118
513
53
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.3.4 Selection of Final Candidate Metrics
We reduced the number of candidate metrics using a series of screening procedures, following
Stoddard et al. (2008). The original (i.e., prior to any adjustment for watershed area) metric response
values were evaluated for range. To evaluate repeatability, we calculated Signal:Noise (S:N) for each
metric following Kaufmann et al. (1999), to compare the variance observed at revisit sites (within
the index period) with the total variance observed across all sites. For adjusted metrics, the S:N
value was calculated after adjusting for watershed area to remove the effects of natural variability
from the "signal", as suggested by Esselman et al. (2013). For both original and adjusted metrics,
the mean response values of the set of least-disturbed reference sites and the set of most-disturbed
sites were compared with two-sample /-tests (assuming unequal variances). Stoddard et al. (2008)
present the advantages of using / values over other statistics as an indicator of metric
responsiveness to disturbance. A candidate metric was not generally considered further if it met any
of the following conditions:
A richness metric (NTAX) had a range < 4
A percentage metrics (PTAX, PIND) had a range < 10%, or had a 90th percentile
value=0
A metric had a S:N value < 1.25
A metric had an absolute value of t < 1.73
The set of least-disturbed validation sites was significantly different (ft < 0.05) from the
set of least-disturbed calibration sites (two sample /-test)
Exceptions were made if there were no metrics in a category that passed all the screens. In these
cases, we chose the metric with the best / value to include in the final set of candidate metrics.
Metrics that passed these screens were then sorted by metric category and /-value. In cases where
the "native only" variant was similar in /-value to the "all species" variant, only one was retained
(usually the all species variant unless there was a sizable difference in the S:N value, and then both
variants were retained in the final list of candidate metrics).
6.3.5 Metric Scoring
We rescaled response values for each of the final suite of metrics to a score ranging between 0 and
10. For "positive" metrics (those having higher values in least-disturbed sites) we used the 5th
percentile of all sites to set the "floor" (below which a score of 0 was assigned), and the 95th
percentile of least-disturbed sites to set the "ceiling" (above which a score of 10 was assigned)
following Stoddard et al. (2008) and as described by Blocksom (2003). For "negative" metrics
(where values were higher in the more disturbed sites), the floor was set at the 5th percentile of
least-disturbed sites, and the ceiling was set at the 95th percentile of all sites. We assigned a score to
response values between the floor and ceiling using linear interpolation.
54
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
We summed the metric scores for each site to derive the fish MMI score. We then multiplied the
fish MMI score by (10/number of metrics) to rescale the score to range between 0 and 100.
6.3.6 Selection of Fin al Fish MMIs
For each of the nine aggregated ecoregions, we used the final list of candidate metrics, and
calculated thousands of candidate fish MMIs based on all possible combinations of the eight
metrics (one from each category), as recommended by Van Sickle (2010). This approach allowed us
to evaluate not only the maximum pairwise correlation among a suite of metrics comprising a fish
MMI, but also the mean pairwise correlation of the suite itself. Indices having low mean
correlations among pairs of metrics may perform better than an index containing component
metrics selected to minimize redundancy based on a maximum allowable correlation coefficient
(Van Sickle 2010).
For each candidate fish MMI, we determined:
1. The F value based on comparing the set of least-disturbed vs. the set of more highly
disturbed sites. We derived a /-value as Vf.
2. The difference between the 25th percentile of the set of least-disturbed sites and the 75th
percentile of the set of more highly disturbed sites. This value (SEPDIFF) is an estimate of
the degree of overlap of the respective boxplots, which has been used to evaluate metric and
index performance (Barbour et al. 1996).
To select the "best" fish MMI from the large number of potential candidates, we excluded any
candidate fish MMIs that had a maximum pairwise correlation of >0.7, or which had a S:N ratio of
<2.5 (Table 6.2). We input the /values and the SEPDIFF values for the remaining candidate fish
MMIs into a principal components analysis. We selected the candidate fish MMI that had the
highest score for the first PCA axis for further evaluation. Combining the values for t and
SEPDIFF into a single PCA axis score provided a simple, objective, and repeatable way to select a
fish MMI that had optimal responsiveness to anthropogenic alteration.
We examined the performance of the component metrics across the range of stream sizes sampled
for NRSA. The potential exists for bias in the fish MMI due to different fish species pools being
available for larger rivers versus smaller streams. Differences across the size range might also result
from the different sampling protocols that were used (wadeable, large wadeable, and boatable). We
used the set of least-disturbed sites to examine patterns in metric response values across Strahler
stream order categories. If one of the component metrics in the "best" fish MMI identified for an
aggregated ecoregion showed a noticeable pattern of either increasing or decreasing response with
Strahler order based on examining boxplots of least-disturbed sites across stream orders, we
selected the fish MMI with the next highest PCA axis score.
Table 6.3 presents the regression equations used to adjust metrics that were included in each of the
55
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
nine regional fish MMIs. The number of adjusted metrics included in a final suite of eight metrics
ranged from two (Southern Plains) to six (Northern Plains). For two aggregated ecoregions (Coastal
Plain and Temperate Plains), the ALIEN metric performed better after adjusting for watershed
area. While it is expected that many richness-based metrics would require adjustment, there are a
fair number of proportional metrics (based on either individuals or taxa) that performed better after
adjustment. This may be due to NRSA including a wider range of stream sizes than many other
MMI development efforts that are based on a smaller set of streams (either smaller or larger).
Table 6.2 Number of final candidate fish multimetric indices (MMIs) calculated from the
final set of passed metrics, before and after screening for maximum pairwise correlation
among metrics and signal:noise ratio.
Number of
Number of
Candidate Fish
Candidate fish
MMIs remaining
Aggregated Ecoregion
MMIs calculated
after screening
Northern Appalachians (NAP)
33,264
9,472
Southern Appalachians (SAP)
36,288
21,976
Coastal Plains (CPL)
9,072
1,494
Southern Plains (SPL)
8,064
2,084
Northern Plains (NPL)
27,648
5,092
Temperate Plains (TPL)
21,600
3,115
Upper Midwest (UMW)
90,720
25,692
Western Mountains (WMT)
84,000
7,120
Xeric West (XER)
32,400
13,220
Table 6.3 Regression equations for adjusting metrics for watershed area. LWSAREA NEW
is the logio-transformed value of watershed area in km2. Only metrics that were included in
the final suite of metrics used to construct one of the nine regional fish MMIs are presented.
Coastal Plain Aggregated Ecoregion (CPL)
ALIENPIND_WS=ALIENPIND-(-0.219734+(0.178533* LWSAREA_NEW));
LOTPIND_WS=LOTPIND-(83.680193+(-5.644243*LWSAREA_NEW));
LITHPIND_WS=LITHPIND-(90.591166+(-21.2575*LWSAREA_NEW));
NAT_TOTLNTAX_WS=NAT_TOTLNTAX-(l 0.929299+(2.873952*LWSAREA_NEW));
TOLRNTAX_WS=TOLRNTAX-(l .831029+(1.559498* LWSAREA_NEW));
Northern Appalachians Aggregated Ecoregion (NAP)
LITHPTAX_WS=LITHPTAX-(91.493806+(-9.389536*LWSAREA_NEW));
NTOLPTAX_WS=NTOLPTAX-(83.244125+(-5.594874*LWSAREA_NEW));
TOLRNTAX_WS=TOLRNTAX-(-O.Q72385+ (1.002947* LWSAREA_NEW));
Northern Plains Aggregated Ecoregion (NPL)
LOTNTAX_WS=LOTNTAX-(0.878392+(1.759Q49*LWSAREA_NEW));
MIGRNTAX_WS=MIGRNTAX-(0.438798+(0.39651*LWSAREA_NEW));
56
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
LITHPIND_WS=LITHPIND-(81.213041+ (-13.064343*LWSAREA_NEW));
NTOLPTAX_WS=NTOLPTAX-(121.656224+(-18.471843*LWSAREA_NEW));
NAT_INTLPIND_WS=NAT_INTLPIND-(84.560234+(-21.788603* LWSAREA_NEW));
NAT_CARNNTAX_WS=NAT_CARNNTAX-(-l .380617+(0.928968* LWSAREA_NEW));
Southern Appalachians Aggregated Ecoregion (SAP)
NAT_CENTNTAX_WS=NAT_CENTNT AX-(-0.017051 +(0.776488* LWSAREA_NEW));
NAT_LITHPIND_WS=NAT_LITHPIND-(85.390153+(-10.818128*LWSAREA_NEW));
INVPIND_WS=INVPIND-(26.04262+(11.423482* LWSAREA_NEW));
Southern Plains Aggregated Ecoregion (SPL)
CYPRPTAX_WS=CYPRPTAX-(45.705777+(-9.448293* LWSAREA_NEW));
NAT_MIGRPTAX_WS=NAT_MIGRPTAX-(-0.604356+(0.532868*LWSAREA_NEW));
Temperate Plains Aggregated Ecoregion (TPL)
ALIENNTAX_WS=ALIENNTAX-(-0.22423+(0.200411*LWSAREA_NEW));
NAT_ICTAPIND_WS=NAT_ICTAPIND-(-Q. 189542+(0.816572* LWSAREA_NEW));
NAT_NT OLNTAX_W S=NAT_NT OLNTAX- (1.946393+ (2.107837*LWSAREA_NEW));
CARNNTAX_WS=CARNNTAX-(-0.005878+(1.292597*LWSAREA_NEW));
Upper Midwest Aggregated Ecoregion (UMW)
INTLLOTNTAX_WS=INTLLOTNTAX-(l .09723+ (0.659379* LWSAREA_NEW));
NTOLNTAX_WS=NTOLNTAX-(2.216995+(2.87Q941*LWSAREA_NEW));
TOLRNTAX_WS=TOLRNTAX-(0.3983Q5+(1.755202* LWSAREA_NEW));
Western Mountains Aggregated Ecoregion (WMT)
INTLLOTPTAX_WS=INTLLOTPTAX-(l 10.962575+(-21,540681*LWSAREA_NEW));
NAT_MIGRPTAX_WS=NAT_MIGRPTAX-(90.991326+(-15.318296*LWSAREA_NEW));
NAT_TOTLNTAX_WS=NAT_TOTLNTAX- (0.748128+ (1.104128* LWSAREA_NEW));
Xeric West Aggregated Ecoregion (XER)
MIGRPTAX_WS=MIGRPTAX-(93.412006+(-20.33135* LWSAREA_NEW));
LITHNTAX_WS=LITHNTAX-(-0.265844+ (1.369981*LWSAREA_NEW));
TOLRNT AX_WS=TOLRNTAX-(-Q. 142977+(0.094138*LWSAREA_NEW));
BENTINVPTAX_WS=BENTINVPTAX-(-5.705387+(9.987192* LWSAREA_NEW));
The following subsections provide information on the performance of each of the metrics that
were used to construct a regional fish MMI. The information includes the floor and ceiling values
that were used to develop a score for each metric (Section 6.3.5).
6.3.6.1 Metric Performance and Scoring: Coastal Plain Aggregated Ecoregion
Table 6.4 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Coastal Plain aggregated ecoregion (CPL). The final suite
57
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
included two negative metrics (the alien and tolerance metrics), and five metrics that were adjusted
for watershed area (Table 6.3). Absolute values of /ranged from 2.05 to 5.04, with only two
metrics having a /-value > 4. Signal to noise ratios ranged from 0.6 to 61.7. The life history metric
(percent of migratory taxa that were intolerant to disturbance) had a low S:N ratio, but it was the
best-performing of any of the life history metrics in this aggregated ecoregion.
6.3.6.2 Metric Performance and Scoring: Northern Appalachians Aggregated Ecoregion
Table 6.5 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Northern Appalachians aggregated ecoregion (NAP). The final
suite included three negative metrics (the alien, tolerance, and trophic metrics), and three metrics
thatwere adjusted for watershed area (Table 6.3). Absolute values of/ranged from 2.40 to 8.39,
with five metrics having a /-value > 4. Signal to noise ratios ranged from 1.9 to 180. The trophic
metric (number of invertivore taxa) did not respond as we expected; it is a negative metric in this
fish MMI, indicating that there were more invertivore species in the set of most-disturbed sites than
in the set of least-disturbed sites. However, fish MMIs that included trophic metrics that responded
as expected did not perform as well as the fish MMI constructed using the metrics in Table 6.5.
The INVNTAX metric was the most responsive trophic metric based on the /-value (Table 6.5)
and had a higher S:N ratio than other trophic metrics with similar /-values.
58
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.4 Performance information of metrics used to construct the fish MMI for the Coastal Plain aggregated ecoregion.
Column name is the field name in the NRSA database.
Scoring Information11
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
% nonnative individuals (adjusted for
Alien
ALIENPIND_WS
watershed area)
-2.88
15.8
NEG
-0.49
14.28
Composition
RBCATONTAX
Number of round-bodied sucker taxa
5.04
2.7
POS
0
3.00
% Lotic individuals (adjusted for watershed
Habitat''
LOTPIND_WS
area)
3.65
7.4
POS
-73.80
30.66
Life History
INTLMIGRPTAX
% of taxa that are migratory and intolerant
to disturbance
2.30
0.6
POS
0
5.88
Reproductive'
LITHPIND_WS
% lithophil individuals (adjusted for
watershed area)
4.71
61.7
POS
-81.39
33.83
Number of native taxa (adjusted for
Richness
NAT_TOTLNTAX_WS
watershed area)
2.60
6.8
POS
-15.49
7.21
Number of tolerant taxa (adjusted for
Tolerance
TOLRNTAX_WS
watershed area)
-2.05
11.8
NEG
-3.60
7.12
Trophic
INVPTAX
% of taxa that are invertivores
3.90
7.1
POS
9.09
68.75
11 Based on comparisons of mean values of least-disturbed and most-disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites.
17 Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor=5th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor=5th percentile of least-disturbed sites (a metric value < floor is assigned a
score of 10), and the ceiling=95th percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic: Occupies flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
59
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.5 Performance information of metrics used to construct the fish MMI for the Northern Appalachians aggregated
ecoregion. Column name is the field name in the NRSA database.
Scoring Information11
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Alien
ALIENNTAX
Number of nonnative taxa
-4.02
1.9
NEG
0
4
Composition
SALMNTAX
Number of taxa in family Salmonidae
6.06
5.4
POS
0
2
Habitat''
NAT_RHEOPIND
% individuals that are native and rheophils
6.37
10.2
POS
0
100
Life History
INTLMIGRPIND
% individuals that are migratory and
intolerant to disturbance
2.40
180
POS
0
8
Reproductiveฎ
LITHPTAX_WS
% of taxa that are lithophils (adjusted for
watershed area)
7.46
14.2
POS
-55.684
23.496
% of taxa that are not tolerant (adjusted
Richness
NTOLPTAX_WS
for watershed area)
3.55
3.8
POS
-39.539
22.490
Number of tolerant taxa (adjusted for
Tolerance
TOLRNTAX_WS
watershed area)
-8.39
3.1
NEG
-1.611
6.853
Trophic
INVNTAX
Number of taxa that are invertivores
-5.54
4.0
NEG
0
9
a Based on comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
" Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95 4 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
60
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.3.6.3 Metric Performance and Scoring: Northern Plains Aggregated Ecoregion
Table 6.6 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Northern Plains aggregated ecoregion (NPL). The final suite
included two negative metrics (the alien and composition metrics), and six metrics that were
adjusted for watershed area (Table 6.3). Absolute values of /ranged from 1.24 to 4.59, with only
one metric having a /-value > 4. Signal to noise ratios ranged from 0.4 to 332. The most responsive
alien metric (number of nonnative taxa) did not meet the criteria for responsiveness or repeatability
(Section 6.3.4). Other alien metrics did not show much response to disturbance, suggesting that the
set of least-disturbed sites in this aggregated ecoregion were similar in terms of percent of
individuals or percent of taxa to the set of most-disturbed sites. Therefore, the number of
nonnative taxa metric, a negative metric, was included to represent the alien metric category, even
though its influence was negligible on the overall MMI for the ecoregion.
6.3.6.4 Metric Performance and Scoring: Southern Appalachians Aggregated Ecoregion
Table 6.7 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Southern Appalachians aggregated ecoregion (SAP). The final
suite included three negative metrics (the composition, life history, and tolerance metrics), and
three metrics that were adjusted for watershed area (Table 6.3). Absolute values of /ranged from
0.58 to 12.15, with seven metrics having a /-value > 4. Signal to noise ratios ranged from 1.5 to
23.4. The best alien metric, percent of taxa that are native, was not very responsive. However, the
percent of native taxa are considered to have a positive influence on a fish assemblage, and thus
was included as the alien metric in the regional Fish MMI.
6.3.6.5 Metric Performance and Scoring: Southern Plains Aggregated Ecoregion
Table 6.8 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Southern Plains aggregated ecoregion (SPL). The final suite
included three negative metrics (the composition, life history, and tolerance metrics), and two
metrics that were adjusted for watershed area (Table 6.3). Absolute values of /ranged from 0.61 to
4.26, with only one metric having a /-value > 4. Signal to noise ratios ranged from 4.6 to 113.4. The
most responsive metrics in three categories (composition, life history, and tolerance) did not meet
the criteria for responsiveness (Section 6.3.4).
61
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.6 Performance information of metrics used to construct the fish MMI for the Northern Plains aggregated ecoregion.
Column name is the field names in the NRSA database.
Metric
Category
Column Name
Description
t-
value*
Signal:
Noise
Value6
Scoring Information17
Direction
of
Response
Floor
Ceiling
Alien
ALIENNTAX
Number of nonnative taxa
1.24
-0.4
NEG
0
2
Composition
NAT_CYPRPIND
% individuals that are native and within
the family Cyprinidae
-2.54
2.3
NEG
0
100
Habitat^
LOTNTAX_WS
Number of lotic taxa (adjusted for
watershed area)
4.59
1.5
POS
-5.045
4.352
Life History
MIGRNTAX_WS
Number of migratory taxa (adjusted for
watershed area)
2.69
0.7
POS
-1.907
1.579
Reproductiveฎ
LITHPIND_WS
% individuals that are lithophils (adjusted
for watershed area)
3.14
6.1
POS
-52.180
53.848
Richness
NTOLPTAX_WS
% taxa that are not tolerant (adjusted for
watershed area)
2.52
6.3
POS
-66.112
29.110
Tolerance
NAT_INTLPIND_WS
% individuals that are native and
intolerant (adjusted for watershed area)
1.82
332.4
POS
-42.369
62.153
Trophic
NAT_CARNNTAX_WS
Number of taxa that are native and
carnivores (adjusted for watershed area)
3.81
1.3
POS
-2.091
1.960
11 Based on comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
17 Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95 4 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
62
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.7 Performance information of metrics used to construct the fish MMI for the Southern Appalachians aggregated
ecoregion. Column name is the field names in the NRSA database.
Scoring Information17
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Alien
NAT_PTAX
% of taxa that are native
-0.58
23.4
POS
80
100
Number of taxa within the family
Centrarchidae (adjusted for watershed
Composition
NAT_CENTNTAX_WS
area)
-4.30
1.5
NEG
-1.535
3.620
% of taxa that are native, benthic and not
Habitat^
NAT_NTOLBENTPTAX
tolerant
9.19
3.9
POS
0
66.670
Number of taxa that are native and
Life History
NAT_MIGRNTAX
migratory
-5.42
2.2
NEG
0
4
% of individuals that are native and
Reproductiveฎ
NAT_LITHPIND_WS
lithophils (adjusted for watershed area)
5.85
7.7
POS
-56.528
28.448
Richness
NTOLPTAX
% of taxa that are not tolerant
8.72
8.7
POS
31.820
100
Tolerance
TOLRPTAX
% of taxa that are tolerant
-12.15
9.7
NEG
0
66.670
% of individuals that are invertivores
Trophic
INVPIND_WS
(adjusted for watershed area)
5.01
8.1
POS
-60.259
38.399
11 Based on comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
" Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95111 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
63
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.8 Performance information of metrics used to construct the fish MMI for the Southern Plains aggregated ecoregion.
Column name is the field names in the NRSA database.
Scoring Information11
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Alien
NAT_PIND
% of individuals that are native
1.98
113.4
POS
67.610
100
Composition
CYPRPTAX_WS
% of taxa within the family Cyprinidae
(adjusted for watershed area)
-1.71
4.8
NEG
-16.787
62.614
Habitat''
RHEOPIND
% of individuals that are rheophils
3.49
51.3
POS
0
92.710
Life History
NAT_MIGRPTAX_WS
% of taxa that are native and migratory
(adjusted for watershed area)
-0.61
6.3
NEG
-1.326
31.490
Reproductiveฎ
LITHNTAX
Number of taxa that are lithophils
4.26
6.9
POS
0
6
Number of taxa that are native and not
Richness
NAT_NTOLNTAX
tolerant
1.85
4.6
POS
1
8
Tolerance
TOLRNTAX
Number of tolerant taxa
-1.54
14.4
NEG
2
14
Trophic
HERBPTAX
% of taxa that are herbivores
3.96
19.0
POS
0
25
11 Based on comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
17 Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95 4 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
64
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.3.6.6 Metric Performance and Scoring: Temperate Plains Aggregated Ecoregion
Table 6.9 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Temperate Plains aggregated ecoregion (TPL). The final suite
included three negative metrics (the alien, composition, and life history metrics), and four metrics
thatwere adjusted for watershed area (Table 6.3). Absolute values of/ranged from 1.69 to 6.96,
with three metrics having a /-value > 4. Signal to noise ratios ranged from 1.2 to 12.6. The most
responsive metric in the life history category did not quite meet the criteria for responsiveness
(Section 6.3.4). The life history metric (number of taxa that are migratory and intolerant) also did
not respond as we expected; it is a negative metric in this fish MMI, indicating that there were more
intolerant migratory species in the set of most-disturbed sites than in the set of least-disturbed sites.
6.3.6.7 Metric Performance and Scoring: Upper Midwest Aggregated Ecoregion
Table 6.10 presents the performance and scoring information for the eight metrics thatwere used
to construct the fish MMI for the Upper Midwest aggregated ecoregion (UMW). The final suite
included only one negative metric (the tolerance metric), and three metrics thatwere adjusted for
watershed area (Table 6.3). Absolute values of/ranged from 0.22 to 5.91, with four metrics having
a /-value > 4. Signal to noise ratios ranged from 2.3 to 12.7. The best alien metric, percent of taxa
that are native, was not very responsive. However, the percent of native taxa are considered to have
a positive influence on a fish assemblage, and thus was included as the alien metric in the regional
Fish MMI
6.3.6.8 Metric Performance and Scoring: Western Mountains Aggregated Ecoregion
Table 6.11 presents the performance and scoring information for the eight metrics thatwere used
to construct the fish MMI for the Western Mountains aggregated ecoregion (WMT). The final suite
included three negative metrics (the composition, tolerance, and trophic metrics), and three metrics
thatwere adjusted for watershed area (Table 6.3). Absolute values of/ranged from 1.45 to 5.56,
with five metrics having a /-value > 4. Signal to noise ratios ranged from 1.3 to 23.2. The most
responsive reproductive metric (% of taxa that are lithophils) did not quite meet the criteria for
responsiveness (Section 6.3.4).
65
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.9 Performance information of metrics used to construct the fish MMI for the Temperate Plains aggregated ecoregion.
Column name is the field names in the NRSA database.
Scoring Information17
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Number of nonnative taxa (adjusted for
Alien
ALIENNTAX_WS
watershed area)
-4.12
2.1
NEG
-0.298
2.045
% of individuals that are native and within
Composition
NAT_ICTAPIND_WS
the family Ictaluridae (adjusted for
watershed area)
-2.94
12.6
NEG
-1.940
17.204
Habitat^
RHEONTAX
Number of taxa that are rheophils
5.02
1.7
POS
0
4
Number of taxa that are migratory and
Life History
INTLMIGRNTAX
intolerant
-1.69
1.2
NEG
0
1
Reproductive'
LITHPIND
% of individuals that are lithophils
2.93
1.4
POS
0
97.520
Number of taxa that are native and not
Richness
NAT_NTOLNTAX_WS
tolerant (adjusted for watershed area)
6.96
2.7
POS
-9.403
4.824
Tolerance
INTLPTAX
% of taxa that are intolerant
3.19
4.2
POS
0
50
Number of taxa that are carnivores
Trophic
CARNNTAX_WS
(adjusted for watershed area)
3.40
2.5
POS
-3.761
2.235
11 Based on comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
" Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95111 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
66
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.10 Performance information of metrics used to construct the fish MMI for the Upper Midwest aggregated ecoregion.
Column name is the field names in the NRSA database.
Scoring Information17
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Alien
NAT_PTAX
% of taxa that are native
-0.22
4.8
POS
85.710
100
Number of taxa within the family
Composition
CYPRNTAX
Cyprinidae
2.19
3.6
POS
0
9
Number of taxa that are lotic and
Habitat^
INTLLOTNTAX_WS
intolerant (adjusted for watershed area)
5.91
2.9
POS
-3.287
2.110
Life History
INTLMIGRPTAX
% of taxa that are migratory and intolerant
3.04
2.3
POS
0
13.330
Reproductiveฎ
LITHPIND
% of individuals that are lithophils
4.22
12.7
POS
0
95.350
Number of taxa that are not tolerant
Richness
NTOLNTAX_WS
(adjusted for watershed area)
4.54
8.2
POS
-8.389
6.445
Number of tolerant taxa (adjusted for
Tolerance
TOLRNTAX_WS
watershed area)
-3.02
3.1
NEG
-3.785
5.549
% of taxa that are invertivores and
Trophic
INTLINVPTAX
intolerant
5.40
7.5
POS
0
33.330
a Based on the comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
17 Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95 4 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
67
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.11 Performance information of metrics used to construct the fish MMI for the Western Mountains aggregated
ecoregion. Column name is the field names in the NRSA database.
Scoring Information11
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Alien
NAT_PIND
% of individuals that are native
5.56
10.5
POS
0
100
NAT CATOPIND
% of individuals that are native and within
-4.20
15.2
NEG
0.000
68
Composition
the family Catastomidae
INTLLOTPTAX WS
% of taxa that are lotic and intolerant
4.23
7.3
POS
-72.045
27.826
Habitat''
(adjusted for watershed area)
NAT_MIGRPTAX_WS
% of taxa that are native and migratory
2.10
23.2
POS
-74.290
40.433
Life History
(adjusted for watershed area)
Reproductive'
LITHPTAX
% of taxa that are lithophils
1.45
16.6
POS
25.000
100
NAT_TOTLNTAX_WS
Number of native taxa (adjusted for
2.39
1.5
POS
-3.009
3.272
Richness
watershed area)
Tolerance
TOLRNTAX
Number of tolerant taxa
-4.71
1.3
NEG
0
2
Trophic
NAT_HERBPTAX
% of taxa that are native and herbivores
-4.20
8.2
NEG
0
33.330
a Based on the comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
" Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95 4 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
68
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.3.6.9 Metric Performance and Scoring: Xeric West Aggregated Ecoregion
Table 6.12 presents the performance and scoring information for the eight metrics that were used
to construct the fish MMI for the Xeric West aggregated ecoregion (XER). The final suite included
two negative metrics (the composition and tolerance metrics), and four metrics that were adjusted
for watershed area (Table 6.3). Absolute values of /ranged from 1.45 to 5.56, with five metrics
having a /-value > 4. Signal to noise ratios ranged from 3.4 to 21.8.
6.4 Fish MMI performance
We evaluated several aspects of performance of the nine regional fish MMIs (Table 6.13). We
compared the fish MMI scores from a set of validation least-disturbed sites to those of the set of
calibration least-disturbed sites to confirm that the models were behaving as anticipated. For all
nine regional fish MMIs, the mean values of the validation sites and sites used to evaluate the
metrics and construct the fish MMIs were not significantly different (two-sample /-test).
We evaluated the responsiveness of the regional fish MMIs to disturbance using two measures: 1) /-
tests to compare the fish MMI scores for the set of least-disturbed sites to those for the set of more
highly disturbed sites (Stoddard et al. 2008), and 2) the difference between the 25th percentile of
least-disturbed sites and the 75th percentile of the set of most-disturbed sites. Boxplots are
presented in Figure 6.3. The results of /-tests (two sample tests assuming unequal variances) and
the percentile differences are presented in Table 6.13. The /-values ranged from 5.71 for the
Northern Plains to 15.38 for the Northern Appalachians. The percentile differences were all
positive (i.e., the boxes did not overlap), and ranged from 0.75 for the Western Mountains to 22.4
for the Northern Appalachians.
We estimated precision of the fish MMIs by calculating the standard deviation of standardized fish
MMI scores (dividing each value by the mean) from all least-disturbed sites. Precision values greater
than zero provide an indication of the remaining disturbance signal left in the set of least-disturbed
sites, plus measurement error. Precision values ranged from 0.10 in the Xeric West aggregated
ecoregion to 0.28 in the Northern Plains and the Upper Midwest aggregated ecoregions. Precision
values between 0.10 and 0.25 are comparable to values obtained for other predictive models of taxa
loss (Hawkins et al. 2010a).
We evaluated the repeatability of the regional fish MMIs using a set of sites that were visited at least
twice during the course of the NRSA 2008-09 project, typically two times in a single year
(Kaufmann et al. 1999, Stoddard et al. 2008). We used a general linear model (PROC GLM, SAS v.
9.12) to obtain estimates of among-site and within-site (from repeat visits) variability. PROC GLM
was used because of the highly unbalanced design (only a small subset of sites had repeat visits). We
used a nested model (sites within year) where both site and year were random effects.
69
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.12 Performance information of metrics used to construct the fish MMI for the Xeric West aggregated ecoregion.
Column name is the field name in the NRSA database.
Scoring Information17
Signal:
Direction
Metric
t-
Noise
of
Category
Column Name
Description
value*
Value6
Response
Floor
Ceiling
Alien
NAT_PIND
% of individuals that are native
7.75
5.4
POS
0
100
CENTPTAX
% of taxa that are within the family
-4.12
3.4
NEG
0
25.000
Composition
Centrarchidae
Habitat''
RHEOPIND
% of individuals that are rheophils
5.03
13.1
POS
0
100
MIGRPTAX_WS
% of taxa that are migratory (adjusted for
1.79
7.9
POS
-64.832
38.279
Life History
watershed area)
LITHNTAX_WS
Number of taxa that are lithophils
8.39
6.4
POS
-6.202
1.649
Reproductive'
(adjusted for watershed area)
Richness
NTOLPTAX
% of taxa that are not tolerant
8.02
8.9
POS
0
100
TOLRNTAX_WS
Number of taxa that are tolerant (adjusted
-8.56
3.7
NEG
-0.129
4.670
Tolerance
for watershed area)
BENTINVPTAX_WS
% of taxa that are benthic invertivores
6.43
21.8
POS
-48.740
23.306
Trophic
(adjusted for watershed area)
a Based on the comparison of mean values of least disturbed and most disturbed sites (validation sites have been excluded).
h Based on variability among sites vs. variability within sites (validation sites have been excluded).
17 Direction: POS=Positive metric (mean value for least-disturbed sites is greater than mean value for most-disturbed sites). NEG=negative
metric (mean value for most-disturbed sites is greater than the mean value for least-disturbed sites). For positive metrics, the floor= 5 th
percentile of all sites (a metric value < floor is assigned a score of 0), and the ceiling=95th percentile of least-disturbed sites (a metric value
> ceiling is assigned a score of 10). For negative metrics, the floor= 5 th percentile of least-disturbed sites (a metric value < floor is assigned
a score of 10), and the ceiling=95 4 percentile of all sites (a metric value > ceiling is assigned a score of 0).
d Habitat metrics: Lotic species occupy flowing water habitats; Rheophils occupy fast water habitats.
e Reproductive metrics: Lithophils require clean substrate for spawning.
70
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.13 Performance statistics for the nine regional fish MMIs.
Coastal
Northern
Northern
Southern
Southern
Temperate
Upper
Western
Xeric
Plain
Appalachians
Plains
Appalachians
Plains
Plains
Midwest
Mountains
West
Performance
Fish
Fish
Fish
Fish
Fish
Fish
Fish
Fish
Fish
Characteristic
MMI
MMI
MMI
MMI
MMI
MMI
MMI
MMI
MMI
Validation least-
*=-1.22
*=1.00
*=1.12
*=0.92
*=-0.02
*=0.41
*=1.40
*=0.43
*=0.86
disturbed sites vs.
least-disturbed
sites used in MMI
development
Least-disturbed
t=10.3
t= 15.38
t-5.71
t-14.7
t=8.07
t-9.76
*=7.45
*=7.74
*=11.42
sites vs. most-
disturbed sites
Difference
+6.2
+22.4
+ 1.6
+8.8
+2.4
+8.5
+ 1.42
+0.75
+9.5
between 25th
percentile of least-
disturbed sites
and 75th percentile
of most-disturbed
sites
Model precision
0.17
0.22
0.28
0.17
0.14
0.15
0.28
0.0.11
0.10
(SD of least-
disturbed sites)
Repeatability
6.6
71.2
4.4
6.5
13.5
6.2
4.3
29.1
9.8
(Signal:Noise)
71
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
COASTAL PLAIN (CPL)
TEMPERATE PLAINS (TPL)
UPPER MIDWEST (UMW)
SOUTHERN APPALACHIANS (SAP)
100
80 ฆ
'r*-.
40
20-
i_rJ
T
' ^ :
WESTERN MOUNTAINS (WMT)
100 -|
SOUTHERN PLAINS (SPL)
NORTHERN PLAINS (NPL)
NORTHERN APPALACHIANS (NAP)
100
XERIC WEST (XER)
Figure 6.3 Boxplots comparing regional fish MMI scores of least-disturbed sites to most-
disturbed sites. Whiskers indicate 10th and 90th percentiles. Points indicate 5th and 95th
percentiles.
72
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
We estimated repeatability by deriving a SignakNoise (S:N) ratio as (F 1)/c, where Fis the F-statistic
from the ANOVA, and c is a coefficient in the equation used to estimate the expected mean square. If
all sites had repeat visits, c would equal 2 (Kaufmann et al. 1999). If no sites had repeat visits, c would
equal 1. SignakNoise ratios ranged from 4.3 in the Upper Midwest aggregated ecoregion to 71.2 in the
Northern Appalachians aggregated ecoregion. High values of S:N need to be interpreted in the
context of the number of repeat visit sites included in the analysis. Artificially high values of S:N can
result if there are a small number of repeat visit sites that have little (or no) variance in fish MMI
scores among them.
We examined the performance of the fish MMIs across the range of stream sizes sampled for NRSA.
The potential exists for bias in the fish MMI due to different fish species pools being available for
larger rivers versus smaller streams. Differences across the size range might also result from the
different sampling protocols that were used according to river or stream size (wadeable, large
wadeable, and boatable). We used the set of least-disturbed sites to examine patterns in fish MMI
scores across three size categories based on Strahler order (Figure 6.4). In most aggregated
ecoregions, there is little difference between the distribution of fish MMI scores among stream size
classes. In the Northern Appalachians aggregated ecoregion, fish MMI scores at the least-disturbed
sites that are 5th order or larger are significantly different from fish MMI scores in least-disturbed sites
that are first or second order (one-way ANOVA with Tukey multiple comparisons test,_p < 0.05).
We examined the potential effect of the three different fish sampling protocols for streams of
different sizes (Figure 6.5). The distribution of fish MMI scores in least-disturbed sites were similar
among the three protocols for most of the nine aggregated ecoregions. In the Northern Appalachians,
this appears to be a tendency for fish MMI scores for least-disturbed sites sampled using the boatable
protocol to be lower than fish MMI scores for least-disturbed sites sampled using either the wadeable
or large wadeable protocols, but the difference is not significantly different (one-way ANOVA with
Tukey multiple comparisons test). In the Northern Plains aggregated ecoregion, fish MMI scores for
least-disturbed sites sampled with the boatable protocol are significantly higher than fish MMI scores
at least-disturbed sites sampled using either the wadeable or large wadeable protocol (one-way
ANOVA with Tukey multiple comparisons test,_p < 0.05). The effect of sampling protocol in the
Upper Midwest and Xeric West aggregated ecoregions are difficult to evaluate, as there was only one
least-disturbed site sampled each using the large wadeable and boatable protocols.
The NRSA includes streams of different temperature regimes as well as a broad range of stream sizes.
We used the predicted summer (July-August) daily stream temperatures (ฐC) based on reference
condition USGS stream temperature stations (Hill et al. 2013) to estimate the mean summer stream
temperature (MSST). We classified the set of least-disturbed streams in each aggregated ecoregion as
either cold water (MSST < 17 ฐC), cool water (MSST between 17 and 20 ฐC) or warm water (MSST >
20 ฐC). Figure 6.6 shows the distribution of fish MMI scores among the three temperature classes for
each aggregated ecoregion. The Coastal Plains aggregated ecoregion did not have any least-disturbed
sites that were classified as either cold or cool water. The Southern Plains aggregated ecoregion did
not have any least disturbed sites classified as cold water (and only two sites classified as
73
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
COASTAL PLAINS (CPL)
100
80
60
40 +
20 -f
0
* /
Strahler Order
SOUTHERN APPALACHIANS (SAP)
u 1UU --
u 80 - -
120
100
80
60
40
20
0
/ f /
J /
Strahler Order
UPPER MIDWEST (UMW)
>? 9
y / ^
#ฆ
Strahler Order
NORTHERN APPALACHIANS (NAP)
100
80
60
40
20
0
1
/ / x
k? 4 $-
Strahler Order
SOUTHERN PLAINS (SPL)
80
60
40
20
0
r
^ J?
^ /
Strahler Order
WESTERN MOUNTAINS (WMT)
100
80
60
40
20
0
1 ^ 5
ffl
o
1? x18"
* /
Strahler Order
NORTHERN PLAINS (NPL)
100
ฃ 80
o
o
ซ 60
ฃ
ฃ 40
.e
il 20
0
y y
Strahler Order
TEMPERATE PLAINS (TPL)
OT 60
Strahler Order
XERIC WEST (XER)
"> 60
Strahler Order
Figure 6.4 Regional fish MMI scores versus Strahler order category (least-disturbed sites).
74
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
COASTAL PLAINS (CPL)
100
ซ 60 ฆ
40
FISH PROTOCOL
NORTHERN APPALACHIANS (NAP)
100
80
60
40
20
0
i #iFN
o
r
: I i
. I O
'
FISH PROTOCOL
WESTERN MOUNTAINS (WMT)
100
80
60
40
20
0
j;
# ^ J
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
COASTAL PLAINS (CPL)
100
80
60
40 - ฆ
20 -:
A ..t
TEMPERATURE CLASS (MSST)
SOUTHERN APPALACHIANS (SAP)
120
100
SO
60
40
20
0
O '.
9 i
I I I
A<> ^
o* ^ J
TEMPERATURE CLASS(MSST)
UPPER MIDWEST (UMW)
100
80
60
40
20
0
a
.o-
/ jr y
^
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
cool water), and the Temperate Plains aggregated ecoregion only had one least-disturbed site classified
as cold water (and only two sites classified as cool water). The Upper Midwest aggregated ecoregion
only had three least-disturbed sites classified as warm water. The Western Mountains aggregated
ecoregion had only one least-disturbed site classified as warm water. In the Northern Appalachians
aggregated ecoregion, fish MMI scores for warm water sites are significantly lower than fish MMI
scores for either cool water or cold water sites (one-way ANOVA with Tukey multiple comparisons
test,_p < 0.001). In the Northern Plains aggregated ecoregion, fish MMI scores for warm water and
cool water sites are significantly lower than fish MMI scores at cold water sites (one-way ANOVA
with Tukey multiple comparisons test,_p < 0.001). In the Xeric West aggregated ecoregion, fish MMI
scores for warm water sites are significantly lower than fish MMI scores for either cool water or cold
water sites (one-way ANOVA with Tukey multiple comparisons test,_p < 0.01).
6.5 Sites WITH LOW FISH ABUNDANCE
The target population of streams and rivers for NRSA includes small perennial headwater streams.
Some very small streams may not contain fish even in the absence of human disturbance. We
followed the approach described by McCormick etal. (2001) and used least-disturbed sites to estimate
a drainage area below which the probability was high that no fish would be present (Table 6.14). This
approach uses the relationship between a set of four physical habitat variables that characterize habitat
volume and the number of fish collected. This relationship defines a habitat volume value below
which nearly all sites sampled were devoid of fish. Then this habitat volume value is related to
watershed area to determine the drainage area below which streams are expected to be naturally
Ashless.
Figure 6.7 shows the results of this analysis. The value for the habitat volume index below which
almost all sites were fishless is 0.41. When habitat volume is plotted against watershed area, this value
corresponds to a watershed area of approximately 2 km2. For sites with watershed areas less than 2
km2 where no fish were collected, we do not report the fish MMI score. Otherwise, we assign a fish
MMI score of zero to sites with no fish collected.
6.6 Benchmarks for assigning ecological condition
For NRSA, ecological condition is based on the deviation from least-disturbed condition (Stoddard et
al. 2006, Hawkins etal. 2010b). Within each of the nine aggregated ecoregions regions, benchmarks
for defining "good" condition and "poor" condition are based on the distribution of fish MMI scores
in least-disturbed sites.
Benchmarks were set following the same process used for benthic macroinvertebrate condition (see
Section 5.3.2). We combined the least-disturbed sites identified for NRSA 2008-09 and 2013-14 to
develop benchmarks that were then applied to the fish MMI scores from both assessments. We used
a single visit per site and used the latest visit if a least-disturbed site was sampled in 2008-09 and
resampled in 2013-14. We attempted to adjust for differences in the quality of least-disturbed sites
across the nine aggregated ecoregions by applying the "hindcasting" approach described in Section
77
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
5.3.2 and by Herlihy etal. (2008), and the NRSA 2008-09 technical report (USEPA 2016).
Table 6.14 Determining the minimum drainage area expected to reliably support the presence
of fish (adapted from McCormick et al (2001)). Variable names are from the NRSA database.
Scores for each metric between the upper and lower criteria were estimated by linear
interpolation.
SET OF SITES
Use least-disturbed sites only (RT_NRSA_FISH="R") to minimize effects of human disturbance
HABITAT VOLUME INDEX
Percent of support reach length that is dry (PCT_DRS)
If PCT_DR < 1%, score = 1. If PCT-DR > 20%, then score = 0.
Logio[(mean wetted width x mean thalweg depth) +0.001] (LXWXD)
If LXWXD > 1, score=l. If LXWXD < -1.4, then score = 0.
Residual pool depth (RP100)
If RP100 > 20, then score=l. If RP100 < 0, then score = 0.
Mean wetted width
If XWIDTH > 6, then score = 1. If XWIDTH = 0, then score = 0.
HABITAT VOLUME INDEX = (PCT_DR score + LXWXD score + RP100 score + XWIDTH
score)/4
PLOT NUMBER OF FISH COLLECTED (TOTLNIND) VS. HABITAT VOLUME INDEX
(QVOLX)
Value for QVOLX below which most sites have no fish = 0.41
PLOT HABITAT VOLUME INDEX VS. WATERSHED AREA (WSAREA_KM2)
QVOLX = 0.41 corresponds to a watershed area of ~ 2 km2
78
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
1000
800
Q)
Q.
E
03
(/)
c
w 600
CO
D
"O
"g 400
ฆ2 200
-
i
-
QVOLX=0.41 #
v '
'
*
.
L_*-|
ฃ>
-ซ #
!*###
0.0
0.2 0.4 0.6 0.8
Habitat Volume Index
1.0
1.0
0.8 --
#
X
CD
"O
c
a) 0.6
E
D
o
15 0.4
+-ป
!5
CO
X
S
..
. a. v
, *
I*
0.41 = 2 km2
0.2 --
0.0
10 20
Watershed Area (km
30 40
2,
50
Figure 6.7 Relationship between number of fish collected, reduced habitat volume, and small
watershed size at least-disturbed sites. Fish are not likely to be found in streams with a
watershed area of < 2 km2. The scales of total number of fish collected and watershed area
axes have been truncated for clarity.
79
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
The benchmarks for each aggregated ecoregion are presented in Table 6.15. The benchmarks for
assigning "good" condition range between > 39.8 for the Upper Midwest to > 76.8 for the Xeric
West. The benchmarks for assigning "poor" condition range from < 29.3 in the Upper Midwest to <
65.4 in the Western Mountains. The hindcasting approach results in the benchmarks in each
aggregated ecoregion differing by 10.5. Note that even though the fish MMI for the Upper Midwest
has lower benchmarks than the other aggregated ecoregions, the fish MMI still performs well (Table
6.13; Figure 6.3).
6.7 Discussion
For NRSA 2008-09, we initially used a model-based approach that adjusted metric responses to
account for natural variability (USEPA 2016; Appendix 7.A) to develop three regional fish MMIs
(Section 6.1.1; Figure 6.1). While these fish MMIs performed adequately for use in the 2008-09
assessment, several constraints limited the ability for users outside of NRSA to make use of that
approach for their data. We evaluated a more traditional approach to developing and evaluating fish
MMIs based on approaches that have evolved from our experience with other regional-scale
assessment efforts (e.g., McCormick et al. 2001, Whittier et al. 2007b, Stoddard et al. 2008, Van Sickle
2010). Using this approach, we constructed a fish MMI that was responsive to disturbance and
repeatable for each of the nine aggregated ecoregions (Table 6.13 and Figure 6.3). The performance
of the nine regional fish MMIs was similar to or better than fish MMIs constructed using random
forest modelling to adjust metric responses for natural variability (Section 6.1.2, Appendix 7.A). Our
evaluation approach focuses on selecting metrics and fish MMIs that maximize responsiveness to
disturbance and have adequate values for other performance criteria such as repeatability. In all nine
aggregated ecoregions, the fish MMIs tend to be more responsive to disturbance and repeatable than
any of their component metrics (Table 6.4 through Table 6.13).
We calculate candidate metrics based on the percent of taxa, which are not commonly considered for
fish. For each of the nine regional fish MMIs, one or more of the final metrics is based on the percent
of taxa (Table 6.4 through Table 6.12). We examine the relationship between watershed area and
metric response for all candidate metrics (not just richness metrics) and adjust the metric response
value when the coefficient of determination (R2) for the linear regression is > 0.10. The fish MMIs for
eight of the nine aggregated ecoregions have at least one metric that is not a richness metric where the
adjusted metric performs better than the unadjusted metric (Table 6.3).
The ability to calculate large numbers of candidate MMIs from a set of metrics that met all our
evaluation criteria is an improvement over stepwise selection of metrics based on correlations with
metrics already selected (Stoddard et al. 2008, Van Sickle 2010). This approach provides the
opportunity to evaluate MMIs based on suites of metrics that might not otherwise be considered and
helps to ensure that the best-performing MMI is selected. Incorporating the difference between the
25th percentile of least-disturbed and 75th percentile of more disturbed sites (Table 6.13) and the F-
score (or /-value) provides a quick and reproducible way of selecting a final fish MMI from the tens of
thousands of candidate fish MMIs that can be generated (Table 6.2). However, within each
80
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 6.15 Benchmarks for assigning ecological condition based on the distribution of
regional fish MMI scores in least-disturbed sites sampled in NRSA 2008-09 or NRSA 2013-14,
adjusted using the hindcasting approach of Herlihy etal. (2008). Aggregated ecoregions are
shown in Figure 6.2. Sample sizes are in parentheses.
Aggregated Ecoregion
Good/Fair
Fair/Poor
Eastern Highlands
Northern Appalachians (60)
> 57.6
< 47.1
Southern Appalachian (94)
> 60.3
< 49.8
Plains and Lowlands
Coastal Plains (39)
> 57.3
< 46.8
Northern Plains (42)
> 46.3
< 35.8
Southern Plains (43)
> 50.2
< 39.7
Temperate Plains (28)
> 58.0
< 47.5
Upper Midwest (28)
> 39.8
< 29.3
West
Western Mountains (70)
> 75.9
< 65.4
Xeric West (25)
> 76.8
< 63.7
aggregated ecoregion, there may be several alternative fish MMIs with similar performance (i.e., a
slightly lower PCA axis score, /-value, and signaknoise ratio) to the fish MMI we selected as the final.
We did note some potential influence of stream size, sampling protocol, and temperature regime on
fish MMI scores in least-disturbed sites in some aggregated ecoregions (Figure 6.4 through Figure
6.6). These patterns were less evident in the original fish MMIs we developed for the three climatic
regions (Figure 6.1; USEPA 2016). The fish MMIs in these aggregated ecoregions still performed well
despite these influences (Figure 6.3, Table 6.13). At the scale of our aggregated ecoregions, small
sample sizes and, in some cases, a limited geographic range of some classes of least-disturbed sites,
make developing separate MMIs for different types of streams (e.g., larger streams or warm water
streams) impractical.
We can consider several future refinements to the NRSA fish MMI development process as data are
acquired from future rounds of NRSA. At present, we cannot develop fish MMIs for those relatively
few NRSA sites that are sampled by seining. These sites tend to be confined to certain geographic
areas. Once we have acquired seining data from enough sites, we may be able to construct a fish MMI
that performs well and is compatible with the fish MMIs developed based on electrofishing data. An
increased pool of least-disturbed sites in each of the nine aggregated ecoregions would allow for a
more rigorous evaluation of the potential influence of factors such as stream size, protocol, and
temperature regime. For larger streams, a national-scale index might be feasible given the advances in
available techniques used to construct and evaluate MMIs. We have the data to construct numeric
tolerance values for individual fish species based on a national-scale data, which would expand upon
previous efforts (Meador and Carlisle 2007, Whittier et al. 2007a) and provide a tool with broad
applicability to bioassessment activities. The fish MMIs we developed are tailored to respond to a
general measure of disturbance, rather than being comprised of metrics that are responsive to different
types of specific stressors. Examining the relationships between metrics and individual stressors would
81
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
improve the interpretability of the fish MMI and the resultant estimates of risk that are produced as
part of the overall assessment in NRSA.
Fish counts, metrics, and multimetric index condition from NRSA are available to download from the
NARS datawebpage - https://www.epa.gov/national-aquatic-resource-surveys/data-national-aquatic-
resource-surveys
6.8 Literature cited
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid bioassessment protocols for
use in streams andwadeable rivers: periphyton, benthic macroinvertebrates, and fish. EPA
841/B-99/002, Office of Water. US Environmental Protection Agency, Washington, DC.
Blocksom, K.A. 2003. A performance comparison of metric scoring methods for a multimetric index
for Mid-Atlantic Highlands streams. Environmental Management 31:0670-0682.
Bramblett, R.G., T.R.Johnson, A.V. Zale, and D.G. Heggem. 2005. Development and evaluation of a
fish assemblage index of biotic integrity for northwestern Great Plains streams. Transactions of
the American Fisheries Society 134:624-640.
Emery, E.B., T.P. Simon, F.H. McCormick, P.L. Angermeier, J.E. DeShon, C.O. Yoder, R.E. Sanders,
W.D. Pearson, G.D. Hickman, R.J. Reash, and J.A. Thomas. 2003. Development of a
multimetric index for assessing the biological condition of the Ohio River. Transactions of the
American Fisheries Society 132:791-808.
Esselman, P.C., D.M. Infante, L. Wang, A.R. Cooper, D. Wieferich, Y.P. Tsang, D.J. Thornbrugh, and
W.W. Taylor. 2013. Regional fish community indicators of landscape disturbance to
catchments of the conterminous United States. Ecological Indicators 26:163-173.
Fausch, K.D., J.R. Karr, and P.R. Yant. 1984. Regional application of an index of biotic integrity based
on stream fish communities. Transactions of the American Fisheries Society 115:39-55.
Frimpong, E.A., and P.L. Angermeier. 2009. Fish traits: a database of ecological and life-history traits
of freshwater fishes of the United States. Fisheries 34:487 - 495.
Goldstein, R.M., and M.R. Meador. 2004. Comparisons of fish species traits from small streams to
large rivers. Transactions of the American Fisheries Society 133:971-983.
Hawkins, C.P., Y. Cao, and B. Roper. 2010a. Method of predicting reference condition biota affects
the performance and interpretation of ecological indices. Freshwater Biology 55:1066-1085.
Hawkins, C.P., J.R. Olson, and R.A. Hill. 2010b. The reference condition: predicting benchmarks for
ecological and water-quality assessments. Journal of the North American Benthological Society 29:312-
343.
Herlihy, A.T., S.G. Paulsen, J.V. Sickle, J.L. Stoddard, C.P. Hawkins, and L.L. Yuan. 2008. Striving for
consistency in a national assessment: the challenges of applying a reference-condition approach
at a continental scale. Journal of the North American Benthological Society 27:860-877.
Hill, R.A., C.P. Hawkins, and D.M. Carlisle. 2013. Predicting thermal reference conditions for USA
streams and rivers. Freshwater Science 32:39-55.
Hughes, R.M., S. Howlin, and P.R. Kaufmann. 2004. A biointegrity index for coldwater streams of
western Oregon and Washington. Transactions of the American Fisheries Society 133:1497-1515.
82
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Karr, J.R. 1981. Assessment of biotic integrity using fish communities. Fisheries 6:21-27.
Kaufmann, P.R., P. Levine, E.G. Robison, C. Seeliger, and D.V. Peck. 1999. Quantifying physical
habitat in wadeable streams. EPA 620/R-99/003, Office of Research and Development, US
Environmental Protection Agency, Washington, DC.
Lyons, J., R.R. Piette, and K.W. Niermeyer. 2001. Development, validation, and application of a fish-
based index of biotic integrity for Wisconsin's large warmwater rivers. Transactions of the
American Fisheries Society 130:1077-1094.
McCormick, F.H., R.M. Hughes, P.R. Kaufmann, D.V. Peck, J.L. Stoddard, and A.T. Herlihy. 2001.
Development of an index of biotic integrity for the Mid-Atlantic Highlands region. Transactions
of the American Fisheries Society 130:857-877.
Meador, M.R., and D.M. Carlisle. 2007. Quantifying tolerance indicator values for common stream fish
species of the United States. Ecological Indicators 7:329338.
Mebane, C.A., T.R. Maret, and R.M. Hughes. 2003. An Index of Biological Integrity (IBI) for Pacific
Northwest Rivers. Transactions of the American Fisheries Society 132:239-261.
Nelson, J.S., E.J. Crossman, H. Espinosa-Perez, L.T. Findley, C.R. Gilbert, R.N. Lea, and J.D.
Williams. 2004. Common and Scientific Names of Fishes from the United States Canada and
Mexico. Sixth edition. Special Publication 29, American Fisheries Society, Bethesda, Maryland.
Oberdorff, T., D. Pont, B. Hugueny, and J.P. Porcher. 2002. Development and validation of a fish-
based index (FBI) for the assessment of "river health" in France. Freshwater Biology 47:1720-
1734.
Page, L.M., and B.M. Burr. 2011. A field guide to freshwater fishes of North America north of Mexico.
Second edition. Houghton Mifflin, Boston, Massachusetts.
Page, L.M., H. Espinosa-Perez, L.T. Findley, C.R. Gilbert, R.N. Lea, N.E. Mandrak, R.L. Mayden, and
J.S. Nelson. 2013. Common and Scientific Names of Fishes from the United States Canada and
Mexico. Seventh edition. Special Publication 34, American Fisheries Society, Bethesda,
Maryland.
Pearson, M.S., T.R. Angradi, D.W. Bolgrien, T.M. Jicha, D.L. Taylor, M.F. Moffett, and B.H. Hill.
2011. Multimetric Fish Indices for Midcontinent (USA) Great Rivers. Transactions of the American
Fisheries Society 140:1547-1564.
Pont, D., R.M. Hughes, T.R. Whittier, and S. Schmutz. 2009. A Predictive Index of Biotic Integrity
Model for Aquatic-Vertebrate Assemblages of Western U.S. Streams. Transactions of the American
Fisheries Society 138:292-305.
Pont, D., B. Hugueny, and C. Rogers. 2007. Development of a fish-based index for the assessment of
river health in Europe: the European Fish Index. Fisheries Management and Ecology 14:427-439.
Roset, N., G. Grenouillet, D. Goffaux, D. Pont, and P. Kestemont. 2007. A review of existing fish
assemblage indicators and methodologies. Fisheries Management and Ecology 14:393-405.
Simon, T.P., and J. Lyons. 1995. Application of the index of biotic integrity to evaluate water resource
integrity in freshwater ecosystems. Pages 245-262 in W. S. Davis and T. P. Simon, editors.
Biological assessment and criteria: tools for water resource planning and decision making. CRC
Press, Boca Raton, Florida.
83
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Stoddard, J.L., A.T. Herlihy, D.V. Peck, R.M. Hughes, T.R. Whittier, and E. Tarquinio. 2008. A
process for creating multimetric indices for large-scale aquatic surveys. Journal of the North
American BenthologicalSociety 27:878-891.
Stoddard, J.L., D.P. Larsen, C.P. Hawkins, R.K. Johnson, and R.H. Norris. 2006. Setting expectations
for the ecological condition of streams: the concept of reference condition. Ideological
Applications 16:1267-1276.
Tejerina-Garro, F.L., B. de Merona, T. Oberdorff, and B. Hugueny. 2006. A fish-based index of large
river quality for French Guiana (South America): method and preliminary results. Aquatic Fining
Resources 19:31-46.
USEPA (United States Environmental Protection Agency). 2009. National Rivers and Streams
Assessment: Field Operations Manual. EPA 841/B-04/004, Office of Water and Office of
Environmental Information, US Environmental Protection Agency, Washington, DC.
USEPA (United States Environmental Protection Agency). 2013a. National Rivers and Streams
Assessment 2013/14: Field Operations Manual -- Wadeable. EPA 841/B-12/009b, Office of
Water and Office of Environmental Information, US Environmental Protection Agency,
Washington, DC.
USEPA (United States Environmental Protection Agency). 2013b. National Rivers and Streams
Assessment 2013/14: Field Operations Manual Non-Wadeable. EPA 841/B-12/009a, Office
of Water and Office of Environmental Information, US Environmental Protection Agency,
Washington, DC.
USEPA (United States Environmental Protection Agency). 2016. National Rivers and Streams
Assessment 2008-2009 Technical Report. EPA 841/R-l 6/008, Office of Water and Office of
Research and Development, US Environmental Protection Agency, Washington, DC.
Van Sickle, J. 2010. Correlated metrics yield multimetric indices with inferior performance. Transactions
of the American Fisheries Society 139:1802-1917.
Yander Laan, J-J-, and C.P. Hawkins. 2014. Enhancing the performance and interpretation of
freshwater biological indices: An application in arid zone streams. Ecological Indicators
36:470-482.
Whittier, T.R., R.M. Hughes, G.A. Lomnicky, and D.V. Peck. 2007a. Fish and amphibian tolerance
values and an assemblage tolerance index for streams and rivers in the western USA.
Transactions of the American Fisheries Society 136:254-271.
Whittier, T.R., R. M. Hughes, J. L. Stoddard, G.A. Lomnicky, D.V. Peck, and A.T. Herlihy. 2007b. A
structured approach for developing indices of biotic integritythree examples from western
streams and rivers in the USA. Transactions of the American Fisheries Society 136:718-735.
84
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Appendix 6.A Comparison of model-based and traditional fish
MULTIMETRIC INDICES FOR NRSA 2008-09
We used the data from the NRSA 2008-09 study to compare the performance of fish MMIs
developed with predictive models (random forests) to adjust metric responses for natural variability
to the performance of fish MMIs developed with a more traditional approach, where metric
responses are adjusted for watershed area using linear regression (Sections 6.1.1 and 6.1.2). The
development and evaluation process for both approaches is essentially the same; the objective
being that the fish MMIs are the best representation of fish ecology that is responsive to
disturbance (Sections 6.4 through 6.6). Both fish MMIs are developed using the same sets of least
disturbed (LD) and most disturbed (MD) sites. Both fish MMIs are comprised of eight metrics and
include a combination of adjusted and unadjusted metrics. Condition classes were developed for
both fish MMIs in the same manner (Section 6.7).
We evaluated four aspects of the performance of the fish MMIs: responsiveness to disturbance,
precision, repeatability, and sensitivity.
6.A.1 Responsiveness to Disturbance and Precision
Figure 6.A.1 compares scores for both types of fish MMIs in least disturbed and most disturbed
sites in each of the nine aggregated ecoregions. The model-based fish MMI was developed for the
three large climatic regions (Figure 6.1), but the fish MMI scores are broken down for each of the
nine aggregated ecoregions. In general, the distributions of least disturbed and most disturbed sites
are similar for both types of fish MMIs.
We used a /-test between least disturbed and most disturbed sites as our performance test for
responsiveness to disturbance (Figure 6.A.2). Sample sizes of least disturbed and most disturbed
sites within each ecoregion were similar if not identical for both types of fish MMI. For both types
of fish MMI, differences between mean values of least disturbed and most disturbed sites were
highly significant in all aggregated ecoregions (p < 0.0001). The traditional fish MMIs had higher
values for t in all but one aggregated ecoregion (Xeric West).
We used the standard deviation of fish MMI scores in least disturbed sites, after adjusting scores by
dividing by the mean value, as our performance test for precision. Precision is expected to be zero
if our adjustments have accounted for natural variability. Precision values greater than zero
represent any disturbance signal remaining after adjustment as well as measurement error. Neither
approach to constructing the fish MMI completely adjusted for natural variability (Figure 6.A.2),
but the amount of unexplained variability in both types of fish MMIs did not impact the ability of
the fish MMIs to be responsive to disturbance (e.g., in the Northern Appalachians aggregated
ecoregion, the traditional fish MMI was comparatively imprecise, but was very responsive to
disturbance). The model-based fish MMIs tended to be slightly more precise than the traditional
MMIs. Both types of fish MMIs were comparatively imprecise in the aggregated ecoregions that
were included in the Plains and Lowlands climatic region.
85
-------
National Riyers and Streams Assessment 2013-2014 Technical Support Document
&EPA
Least Disturbed (LD) Sites vs.
Most Disturbed (MD) Sites
RF= model-based
Trad=Traditional
I1= LP
I1= MD
WMT
<2 25
XER
100-
s 50
& 25
RF
Trad
UMW
o 75
*
2 25
,S2 25
2 50
<2 25
7/12/2017
8 75
00
1 50
2
!ง 25
u-
0
RF
Trad
Steering Commi
CPL
SAP
100
0)
8 75
00
1 50
2
* 25
u.
0
100-
0)
8 75
00
1 50-
2
ป 25
u.
0
RF Trad
RF Trad
Figure 6.A.1 Distribution of fish MMI scores in least disturbed vs most disturbed sites in NRSA 2008-09. For each aggregated ecoregion (see
Figure 6.2), the left-hand pair of boxplots are for the model-based fish MMI (RF), and the right-hand pair are for the "traditional" fish MMI
(Trad). Gray boxes=least-disturbed sites and red boxes=most-disturbed sites.
86
-------
National Riyers and Streams Assessment 2013-2014 Technical Support Document
&EPA
Responsiveness and
Precision
Responsiveness (f-value)
X)
TO
V
3
TO
>
15 --
10 --
5 --
-1iIIIIi pI IIIIII\~7
NAP II
/ --
~ D
~'
UMW Q
NPL Q
+
+
+
0 5 10 15
f-value (Random Forest FMMI)
(/>
3 0.30
in
3 ง 0.25
TS 5
ฃ if
ป ฆ= 0.20
f ฃ 0.15
a>
P5
tO.10
o
w 005
Precision
i I i i i i I i i i i I
NPL
NAP 7_L
~ UMW U,'
ns
x
+
m
Ph D
| 1 1 ฆ ฆ | 1 1 1 1 |
0.05 0.10 0.15 0.20 0.25 0.30
SD of Mean-adjusted LD Sites
(Random Forest FMMI)
Traditional FMMIs have higher f-
values
All are significant at p < 0.0001 for
both FMMIs
7/12/2017
Modeled FMMIsslightly more precise
FMMIsin Plainsand Lowlandsecoregions
have poorest precision
Still a fair amount of unexplained variability
in both types of FMMIs, but it seems to have
little impact on responsiveness
18
NRSASteenngCommittee Webmar
Figure 6.A.2 Comparison of two types of fish MMI scores for responsiveness to disturbance and precision in nine aggregated ecoregions (see
Figure 6.2). Y-axis= Traditional fish MMI score; X-axis=model-based fish MMI score; line is a 1:1 line. Colors coincide with regions on map
inset (violet=Eastern Highlands, brown=Plains and Lowlands, blue=Western Mountains and Xeric).
87
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.A.2 Repeatability and Sensitivity
We evaluated repeatability of the fish MMIs by calculating a SignakNoise ratio (S:N; see Section 6.4),
which compared the variance among sites to variance within sites (from repeat visits). We estimated
the sensitivity of the fish MMIs based on the proportion of most disturbed sites that were
significantly different (using an interval test) from the set of least disturbed sites. The interval test is
more conservative than simply looking at the number of most disturbed sites that are below a single
percentile value (e.g., the 5th percentile) of the least disturbed sites.
Figure 6.A.3 shows the results of the comparisons for repeatability and sensitivity. S:N ratios for
both types of fish MMIs are adequate for use in NRSA. The model-based fish MMIs tend to have
slightly higher values of S:N than the traditional fish MMIs. In some aggregated ecoregions (e.g., the
Northern Appalachians or the Upper Midwest), there is a small number of sites with repeat visits. If
there is little or no variability in the fish MMI scores at these sites between visits, it will result in a
very high estimate of S:N that is mostly a function of small sample size.
Sensitivity values are similar for both types of fish MMIs and are nearly identical for those
aggregated ecoregions in the Plains and Lowlands climatic regions. The low values (< 40% for all
but one aggregated ecoregion) reflect the variability present in both the least disturbed and most
disturbed sites at the scale of the aggregated ecoregions. The lowest sensitivity was seen in the
Northern Plains aggregated ecoregion, while the greatest sensitivity was observed in the Xeric West
aggregated ecoregion.
6.A.3 Correlation of Fish MMI Scores
We looked at how similar the traditional fish MMI scores were to the model-based fish MMI scores
at all sites. For each aggregated ecoregion, we calculated the Pearson correlation coefficient, and
calculated the geometric mean functional regression (GMFR) because each fish MMI is measured
with error. We used a single index visit for each site and excluded sites where no fish were collected.
Correlation coefficients are > 0.7 for all but the Northern Plains aggregated ecoregion (Figure
6.A.4). The GFMR analysis indicates that for two aggregated ecoregions (the Upper Midwest and
the Western Mountains), the two fish MMIs are identical (slope=l, intercept=0). In the Xeric West
aggregated ecoregion, the traditional fish MMI scores are consistently higher by a small amount than
the model-based fish MMI scores (slope=l, intercept > 0). For the remaining aggregated ecoregions,
slopes are >1 except in the Southern Plains aggregated ecoregion.
88
-------
National Riyers and Streams Assessment 2013-2014 Technical Support Document
SEPA
Repeatability
and Sensitivity
Repeatability
80
2 60
o
.2 40
o 20 --
.ฎ 0
to
-iir-
-ir
NAP
WMT
~
a rr
^CT ฃ?
j-iT i | i i i i | < i
i i I
+
UMW
i i
0 5 10 15 20
Signal:Noise fRandom Forest FMMI)
80
a
-i
.E ! 60 -
0 IL
z ซ
= 40 +
A) o
S*
1 2 20"
5 i-
0
Sensitivity
+
~XER-
NpL B
&
A
-ป'I11h
+
20 40 60 80
% MD Sites Not In LD
(Random Forest FMMI}
Modeled FMMIs slightly more
repeatable
Affected by number of repeat visits
(little or no within site variance would
inflate the S:N valueO
Sensitivity is similar for PLNLOW
regions
Traditional MMI is slightly more
sensitive in EHIGH, RF slightly better
in WMTNS regions
7/12/2017
N RSA Steeri ng Com mittee Webi nar
Figure 6.A.3 Comparison of two types of fish MMI scores for repeatability and sensitivity in nine aggregated ecoregions (see Figure 6.2). Y-axis1
Traditional fish MMI score; X-axis=model-based fish MMI score; line is a 1:1 line. Colors coincide with regions on map inset (violet - Eastern
Highlands, brown= Plains and Lowlands, blue=Western Mountains and Xeric).
89
-------
National Riyers and Streams Assessment 2013-2014 Technical Support Document
Random Forest MM I
leringCom
Random Fores! MM I
Random Forest MM)
Random Forest MMI
r=0.76
ฆb
0 25 50 75 100
Random Forest MMI
WMT
r=0.8
0 25 50 75 100
Random Forest MMI
&EPA
FMMI Scores Are Correlated
7/12/2017
Random Forest MMI
UMW
25 SO 75 100
Random Foresi MMI
Random Forest MMI
Figure 6.A.4 Comparison of two types of fish MMI scores in nine aggregated ecoregions (see Figure 6.2). Points are index visits only, and sites
where no fish were collected are excluded. Y-axis= Traditional fish MMI score; X-axis=model-based fish MMI score. Solid black line: 1:1 line;
Dashed Blue line: Regression line based on geometric mean functional regression.
90
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
6.A.4 Population Estimates
The survey design for NRSA allows us to make inferences from the set of sampled sites to a much
larger target population. We wanted to know if the two types of fish MMIs would yield different
estimates of biological condition for the target population. We assigned condition classes for each
type of fish MMI in each aggregated ecoregion using the approach described in Section 6.5.
Condition class is assigned for each of the nine aggregated ecoregions based on the deviation from
the set of LD sites. For the four aggregated ecoregions in the Eastern Highlands and Western
Mountains climatic regions, the two types of fish MMIs yield similar estimates of the percent of
stream length in both good and poor condition (Figure 6.A.5); the largest differences in length are
in the Xeric West aggregated ecoregion. In the Plains and Lowlands climatic region, the two types of
fish MMIs yield similar estimates of the percent of stream length in good and poor condition for all
aggregated ecoregions except for the Coastal Plains, where the traditional fish MMI produces a
smaller percent of stream length in good condition and a larger percent of stream length in poor
condition compared to the model-based fish MMI. One or both types of fish MMIs in the Northern
Plains and the Upper Midwest aggregated ecoregions had some performance issues, yet the
condition class estimates for the sampled target population were similar.
Based on our evaluations, both types of fish MMIs generally have similar performance and provide
similar estimates of biological condition for the samples target population in each aggregated
ecoregion. Scores for the two types of fish MMIs are well correlated despite differences in
component metrics and the scale at which metric adjustments are made (climatic region for the
model-based fish MMI and aggregated ecoregion for the traditional fish MMI). The quality of least-
disturbed sites may be less similar among the five aggregated ecoregions that are included in the
Plains and Lowlands climatic region than the aggregated ecoregions that are included in either the
Eastern Highlands or Western Mountains climatic regions.
We calculated the traditional fish MMI scores for all sites in the NRSA 2008-09 and the NRSA
2013-14. The scale of traditional fish MMI development (i.e., the nine aggregated ecoregions) is
consistent with the MMI developed for the benthic macroinvertebrate assemblage in NRSA. We did
not use any least-disturbed sites identified in NRSA 2013-14 to develop the fish MMIs, but we did
pool the least-disturbed sites from both studies to estimate the benchmark values to assign biological
condition.
91
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
oEPA Extent Estimates are Similar
Eastern Highlands (EHIGH)
ฃ> 20
EC09 Subpopulation
Western Mountains and Xeric (WEST)
WMT XER
EC09 Subpopulation
ฆฆ RF GOOD E=l TR GOOD
RF POOR TR POOR
GOOD= similar to LD sites POOR= different from LD sites
NAP
SAP
WMT
XER
Est. stream length (km) 189,488
509,319 243,444 72,330
AEPA
Coastal Plains Estimates are
Less Similar
(CPL), Southern (SPL), and Northern (NPL) Plains
Coastal
CPL SPL NPL
ECQ9 Subpopulation
Temperate Plains (TPL) and Upper Midwest (UMW)
TPL UMW
ECQ9 Subpopulation
[=l RF GOOD [=1 TR GOOD 3^3 RF POOR [=1 TR POOR
CPL
SPL
NPL
TPL
UMW
Est. stream length (km)
284,065
58,853
43,432
371,316
148,951
NRSA Steering Committee Webinar 22
Figure 6.A.5 Biological condition in nine aggregated ecoregions (EC09) (see Figure 6.2) for
the NRSA 2008-09 based on two types of fish MMIs. Top panel shows aggregated
ecoregions within the Eastern Highlands and Western Mountains climatic regions. Lower
panel shows aggregated ecoregions within the Plains and Lowlands climatic regions. Bars
represent the percent of the length of the sampled target population inferred from the set of
sampled sites; error bars are 95% confidence intervals. RF=model-based fish MMI;
TRtraditional fish MMI. Goodsimilar to least-disturbed sites; Poor=different from least-
disturbed sites (see Section 6.5 and Table 6.15). The total estimated length of the sampled
target population for each aggregated ecoregion is shown in the table.
92
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Appendix 6.B Candidate metrics considered for fish MMI
DEVELOPMENT
Table 6.B.1 presents the candidate metrics that were evaluated for potential inclusions in the regional
fish MMIs. Metric classes represent different attributes of fish assemblage structure or function. Some
metrics are combinations of two different metric classes. Composition metrics generally focus on the
family level of taxonomic resolution. Round-bodied suckers include the following genera: Catostomus,
Cycleptus, Erimy^on, Hypentelium, Minytrema, Moxostoma, Pantosteus, and Thoburnia. Migratory species
include both diadromous and anadromous species. "Not tolerant" metrics include all species not
classified as tolerant (and thus include intolerant species, moderately tolerant species, and species that
have no tolerance category assigned). We use "intolerant" here in the same context as others use
"sensitive."
Table 6.B.1 List of candidate metrics
METRIC VARIABLE
METRIC CLASS
NAME
DESCRIPTION
ALIEN
ALIENNTAX
No. Non-native species
ALIEN
ALIENPIND
% Non-native individuals
ALIEN
ALIENPTAX
% Non-native taxa
ALIEN
NAT_PIND
% Native individuals
ALIEN
NAT_PTAX
% Native taxa
COMPOSITION
CATONTAX
No. Catostomid species
COMPOSITION
CATOPIND
% Catostomid individuals
COMPOSITION
CATOPTAX
% Catostomid taxa
COMPOSITION
NAT_CATONTAX
No. Native catostomid species
COMPOSITION
NAT_CATOPIND
% Native catostomid
individuals
COMPOSITION
NAT_CATOPTAX
% Native catostomid taxa
COMPOSITION
RBCATONTAX
No. Round-bodied catostomid
species
COMPOSITION
RBCATOPIND
% Round-bodied catostomid
individuals
COMPOSITION
RBCATOPTAX
% Round-bodied catostomid
taxa
COMPOSITION
NAT_RBCATONTAX
No. Native round-bodied
catostomid species
COMPOSITION
NAT_RBCATOPIND
% Native round-bodied
catostomid individuals
COMPOSITION
NAT_RBCATOPTAX
% Native round-bodied
catostomid taxa
COMPOSITION
CENTNTAX
No. Centrarchid species (excl.
Micropterus spp.)
COMPOSITION
CENTPIND
% Centrarchid individuals (excl.
Micropterus spp.)
93
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
METRIC VARIABLE
METRIC CLASS
NAME
DESCRIPTION
COMPOSITION
CENTPTAX
% Centrarchid taxa (excl.
Micropterus spp.)
COMPOSITION
NAT_CENTNTAX
No. Native centrarchid species
(excl. Micropterus spp.)
COMPOSITION
NAT_CENTPIND
% native centrarchid
individuals (excl. Micropterus
spp.)
COMPOSITION
NAT_CENTPTAX
% Native centrarchid taxa (excl.
Micropterus spp.)
COMPOSITION
CYPRNTAX
No. Cyprinid species (excluding
all carps and goldfish)
COMPOSITION
CYPRPIND
% Cyprinid individuals
(excluding all carps and
goldfish)
COMPOSITION
CYPRPTAX
% Cyprinid individuals
(excluding all carps and
goldfish)
COMPOSITION
NAT_CYPRNTAX
No. Native cyprinid species
(excluding all carps and
goldfish)
COMPOSITION
NAT_CYPRPIND
% Native cyprinid individuals
(excluding all carps and
goldfish)
COMPOSITION
NAT_CYPRPTAX
% Native cyprinid individuals
(excluding all carps and
goldfish)
COMPOSITION
ICTANTAX
No. Ictalurid species
COMPOSITION
ICTAPIND
% Ictalurid individuals
COMPOSITION
ICTAPTAX
% Ictalurid taxa
COMPOSITION
NAT_ICTANTAX
No. Native ictalurid species
COMPOSITION
NAT_ICTAPIND
% Native ictalurid individuals
COMPOSITION
NAT_ICTAPTAX
% Native ictalurid taxa
COMPOSITION
SALMNTAX
No. Salmonid species
COMPOSITION
SALMPIND
% Salmonid individuals
COMPOSITION
SALMPTAX
% Salmonid taxa
COMPOSITION
NAT_SALMNTAX
No. Native salmonid species
COMPOSITION
NAT_SALMPIND
% Native salmonid individuals
COMPOSITION
NAT_SALMPTAX
% Native salmonid taxa
HABITAT
COLDNTAX
No. Coldwater species
HABITAT
COLDPIND
% Coldwater individuals
HABITAT
COLDPTAX
% Coldwater taxa
HABITAT
NAT_COLDNTAX
No. Native coldwater species
HABITAT
NAT_COLDPIND
% Native coldwater individuals
HABITAT
NAT_COLDPTAX
% Native coldwater taxa
HABITAT
LOTNTAX
No. Lotic species
94
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
METRIC CLASS
METRIC VARIABLE
NAME
DESCRIPTION
HABITAT
LOTPIND
% Lotic individuals
HABITAT
LOTPTAX
% Lotic taxa
HABITAT
NAT_LOTNTAX
No. Native lotic species
HABITAT
NAT_LOTPIND
% Native lotic individuals
HABITAT
NAT_LOTPTAX
% Native lotic taxa
HABITAT
NAT_NTOLBENTNTAX
No. of Native not tolerant
benthic species (BPJ based
tolerance assignments)
HABITAT
NAT_NTOLBENTPIND
% Native not tolerant benthic
individuals (BPJ based
tolerance assignments)
HABITAT
NAT_NTOLBENTPTAX
% Native not tolerant benthic
taxa (BPJ based tolerance
assignments)
HABITAT
RHEONTAX
No. Rheophilic species
HABITAT
RHEOPIND
% Rheophilic individuals
HABITAT
RHEOPTAX
% Rheophilic taxa
HABITAT
NAT_RHEONTAX
No. Native rheophilic species
HABITAT
NAT_RHEOPIND
% Native rheophilic individuals
HABITAT
NAT_RHEOPTAX
% Native rheophilic taxa
HABITAT
NTOLBENTNTAX
No. Not tolerant benthic
species
HABITAT
NTOLBENTPIND
% Not tolerant benthic species
HABITAT
NTOLBENTPTAX
% not tolerant benthic taxa
HABITAT (TOLERANCE)
INTLLOTNTAX
No. Intolerant lotic species
HABITAT (TOLERANCE)
INTLLOTPIND
% Intolerant lotic individuals
HABITAT (TOLERANCE)
INTLLOTPTAX
% Intolerant lotic taxa
HABITAT (TOLERANCE)
NAT_INTLLOTNTAX
No. Native intolerant lotic
species
HABITAT (TOLERANCE)
NAT_INTLLOTPIND
% Native intolerant lotic
individuals
HABITAT (TOLERANCE)
NAT_INTLLOTPTAX
% Native intolerant lotic taxa
HABITAT (TOLERANCE)
INTLRHEONTAX
No. Intolerant rheophilic
species
HABITAT (TOLERANCE)
INTLRHEOPIND
% Intolerant rheophilic
individuals
HABITAT (TOLERANCE)
INTLRHEOPTAX
% Intolerant rheophilic taxa
HABITAT (TOLERANCE)
NAT_INTLRHEONTAX
No. Native intolerant
rheophilic species
HABITAT (TOLERANCE)
NAT_INTLRHEOPIND
% Native intolerant rheophilic
individuals
HABITAT (TOLERANCE)
NAT_INTLRHEOPTAX
% Native intolerant rheophilic
taxa
LIFE HISTORY
(MIGRATION STRATEGY)
MIGRNTAX
No. Migratory species
95
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
METRIC CLASS
METRIC VARIABLE
NAME
DESCRIPTION
LIFE HISTORY
(MIGRATION STRATEGY)
MIGRPIND
% Migratory individuals
LIFE HISTORY
(MIGRATION STRATEGY)
MIGRPTAX
% Migratory taxa
LIFE HISTORY
(MIGRATION STRATEGY)
NAT_MIGRNTAX
No. Native migratory species
LIFE HISTORY
(MIGRATION STRATEGY)
NATJVIIGRPIND
% Native migratory individuals
LIFE HISTORY
(MIGRATION STRATEGY)
NAT_MIGRPTAX
% Native migratory taxa
LIFE HISTORY
(TOLERANCE)
INTLMIGRNTAX
No. Intolerant migratory
species
LIFE HISTORY
(TOLERANCE)
INTLMIGRPIND
% Intolerant migratory
individuals
LIFE HISTORY
(TOLERANCE)
INTLMIGRPTAX
% Intolerant migratory taxa
LIFE HISTORY
(TOLERANCE)
NAT_INTLMIGRNTAX
No. Native intolerant migratory
species
LIFE HISTORY
(TOLERANCE)
NAT_INTLMIGRPIND
% Native intolerant migratory
individuals
LIFE HISTORY
(TOLERANCE)
NAT_INTLMIGRPTAX
% Native intolerant migratory
taxa
REPRODUCTIVE
LITHNTAX
No. Lithophilic spawner
species
REPRODUCTIVE
LITHPIND
% Lithophilic spawner
individuals
REPRODUCTIVE
LITHPTAX
% Lithophilic spawner taxa
REPRODUCTIVE
NAT_LITHNTAX
No. Native lithophilic spawner
species
REPRODUCTIVE
NAT_LITHPIND
% Native lithophilic spawner
individuals
REPRODUCTIVE
NAT_LITHPTAX
% Native lithophilic spawner
taxa
RICHNESS
TOTLNTAX
Total no. distinct species
collected
RICHNESS
NAT_TOTLNTAX
No. Native distinct species
collected
RICHNESS
NTOLNTAX
No. Not tolerant species
RICHNESS
NTOLPIND
% Not tolerant individuals
RICHNESS
NTOLPTAX
% Not tolerant taxa
RICHNESS
NAT_NTOLNTAX
No. Native not tolerant species
RICHNESS
NAT_NTOLPIND
% Native not tolerant
individuals
RICHNESS
NAT_NTOLPTAX
% Native not tolerant taxa
TOLERANCE
INTLNTAX
No. Intolerant species (BPJ-
based tolerance assignments)
96
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
METRIC VARIABLE
METRIC CLASS
NAME
DESCRIPTION
TOLERANCE
INTLPIND
% Intolerant individuals (BPJ-
based tolerance assignments)
TOLERANCE
INTLPTAX
% Intolerant taxa (BPJ-based
tolerance assignments)
TOLERANCE
NAT_INTLNTAX
No. Native intolerant species
(BPJ-based tolerance
assignments)
TOLERANCE
NAT_INTLPIND
% Native intolerant individuals
(BPJ-based tolerance
assignments)
TOLERANCE
NAT_INTLPTAX
% Native intolerant taxa (BPJ-
based tolerance assignments)
TOLERANCE
NAT_TOLRNTAX
No. Native tolerant species
(BPJ-based tolerance
assignments)
TOLERANCE
NAT_TOLRPIND
% Native tolerant individuals
(BPJ-based tolerance
assignments)
TOLERANCE
NAT_TOLRPTAX
% Native tolerant taxa (BPJ-
based tolerance assignments)
TOLERANCE
TOLRNTAX
No. Tolerant species (BPJ-
based tolerance assignments)
TOLERANCE
TOLRPIND
% Tolerant individuals (BPJ-
based tolerance assignments)
TOLERANCE
TOLRPTAX
% Tolerant taxa (BPJ-based
tolerance assignments)
TROPHIC
CARNNTAX
No. Carnivore species
TROPHIC
CARNPIND
% Carnivore individuals
TROPHIC
CARNPTAX
% Carnivore taxa
TROPHIC
NAT_CARNNTAX
No. Native carnivore species
TROPHIC
NAT_CARNPIND
% Native carnivore individuals
TROPHIC
NAT_CARNPTAX
% Native carnivore taxa
TROPHIC
NTOLCARNNTAX
No. Not tolerant carnivore
species
TROPHIC
NTOLCARNPIND
% Not tolerant carnivore
individuals
TROPHIC
NT OLC ARNPTAX
% Not tolerant carnivore taxa
TROPHIC
NAT_NTOLCARNNTAX
No. Native not tolerant
carnivore species
TROPHIC
NAT_NTOLCARNPIND
% Native not tolerant carnivore
individuals
TROPHIC
NAT_NTOLCARNPTAX
% Native not tolerant carnivore
taxa
TROPHIC
HERBNTAX
No. Herbivore species
TROPHIC
HERBPIND
% Herbivore individuals
TROPHIC
HERBPTAX
% Herbivore taxa
97
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
METRIC CLASS
METRIC VARIABLE
NAME
DESCRIPTION
TROPHIC
NAT_HERBNTAX
No. Native herbivore species
TROPHIC
NAT_HERBPIND
% Native herbivore individuals
TROPHIC
NAT_HERBPTAX
% Native herbivore taxa
TROPHIC
INVNTAX
No. Invertivore species
TROPHIC
INVPIND
% Invertivore individuals
TROPHIC
INVPTAX
% Invertivore taxa
TROPHIC
NAT_INVNTAX
No. Native invertivore species
TROPHIC
NAT_INVPIND
% Native invertivore
individuals
TROPHIC
NAT_INVPTAX
% Native invertivore taxa
TROPHIC
NTOLINVNTAX
No. Not tolerant invertivore
species
TROPHIC
NTOLINVPIND
% Not tolerant invertivore
individuals
TROPHIC
NTOLINVPTAX
% Not tolerant invertivore taxa
TROPHIC
NAT_NTOLINVNTAX
No. Native not tolerant
invertivore species
TROPHIC
NAT_NTOLINVPIND
% Native not tolerant
invertivore individuals
TROPHIC
NAT_NTOLINVPTAX
% Native not tolerant
invertivore taxa
TROPHIC
OMNINTAX
No. Omnivore species
TROPHIC
OMNIPIND
% Omnivore individuals
TROPHIC
OMNIPTAX
% Omnivore taxa
TROPHIC
NAT_OMNINTAX
No. Native omnivore species
TROPHIC
NAT_OMNIPIND
% Native omnivore individuals
TROPHIC
NAT_OMNIPTAX
% Native omnivore taxa
TROPHIC (HABITAT)
BENTINYNTAX
No. Benthic invertivore species
TROPHIC (HABITAT)
BENTINVPIND
% Benthic invertivore
individuals
TROPHIC (HABITAT)
BENTINVPTAX
% benthic invertivore taxa
TROPHIC (HABITAT)
NAT_BENTINVNTAX
No. Native benthic invertivore
species
TROPHIC (HABITAT)
NAT_BENTINVPIND
% Native benthic invertivore
individuals
TROPHIC (HABITAT)
NAT_BENTINVPTAX
% Native benthic invertivore
taxa
TROPHIC (HABITAT)
INTLINVNTAX
No. Intolerant invertivore
species
TROPHIC (HABITAT)
INTLINVPIND
% Intolerant invertivore species
TROPHIC (HABITAT)
INTLINVPTAX
% Intolerant invertivore taxa
TROPHIC (HABITAT)
NAT_INTLINVNTAX
No. Native intolerant
invertivore species
TROPHIC (HABITAT)
NAT_INTLINVPIND
% Native intolerant invertivore
species
98
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
METRIC CLASS
METRIC VARIABLE
NAME
DESCRIPTION
TROPHIC (HABITAT)
NAT_INTLINVPTAX
% Native intolerant invertivore
taxa
99
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
7 Water Chemistry Analyses
Water samples were collected as a grab sample from the X site at the midpoint of the reach in
wadeable systems and at Transect A in boatable systems (see NRSA 2013-14 Field Operations
Manual and Laboratory Operations Manual for additional details). The main report presents
assessments for four chemical stressors: total nitrogen (TN), total phosphorus (TP), acidity, and
salinity. These benchmark values and class definitions were identical to those used in the NRSA
2008-09. Water chemistry data, including additional parameters not assessed in the report, are
available to download from the NARS data webpage - https://www.epa.gov/national-aquatic-
resource-surveys/data-national-aquatic-resource-surveys
7.1 Acidity And Salinity Benchmarks
For acidity, criteria values were determined based on values derived during the National Acid
Precipitation Assessment Program (Baker et al. 1990; Kaufmann et al. 1991). Sites with acid
neutralizing capacity (ANC) less than zero were considered acidic. Acidic sites with dissolved
organic carbon (DOC) greater than 10 mg/L were classified as organically acidic (natural). Acidic
sites with DOC less than 10 and sulfate less than 300 |j,eq/L were classified as acidic deposition
impacted, while those with sulfate above 300 |j,eq/L were considered acid mine drainage
impacted. Sites with ANC between 0 and 25 |j,eq/L and DOC less than 10 mg/L were
considered acidic-deposition-influenced but not currently acidic. These low ANC sites typically
become acidic during high flow events (episodic acidity).
Salinity data values were divided into good, fair, or poor classes. Salinity classes were defined
by specific conductance using ecoregional specific values (Table 7.1).
7.2 Total Phosphorus And Total Nitrogen Benchmarks
Total nitrogen and phosphorus concentrations were classified as "good", "fair," or "poor" using
a method similar to that used for macroinvertebrate IBI classes using deviation from reference
site distribution percentiles by aggregate ecoregion (see Herlihy and Sifneos, 2008 for details).
For nutrients, the value at (and below) the 75th percentile of the reference distribution was used
for each ecoregion to define the least-disturbed condition class (goodfair boundary). The 95th
percentile (and above) of the reference distribution in each ecoregion defines the most disturbed
condition class (fair-poor boundary) (Table 7.1).
A set of "nutrient reference sites" was defined for this analysis using both WSA and NRSA data.
All available WSA and NRSA sample sites were screened for water chemical and physical habitat
disturbances using the process described in Chapter 4 with the exception that total phosphorus
and total nitrogen values were not used as screens to avoid circularity in defining nutrient
100
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
benchmarks. Sites with screening values exceeding criteria for the remaining parameters in Table
4.2 were excluded as nutrient reference sites.
To adjust the process after the removal of the nutrient screens, we incorporated screens for land
cover disturbance. A single national criterion was used to exclude sites that had watershed
%Urban LULC (Land Use Land Cover) >10%, watershed road density > 3 km/km2, and
watershed population density >100 people/km2. For watershed %Agriculture LULC screening,
ecoregional specific criteria were used as screens; NAP, WMT, XER (>10%), CPL, NPL, SAP,
SPL, UMW (>25%), TPL (>50%). Before calculating ecoregional nutrient reference site
percentiles, outliers (values outside 1.5 times the interquartile range above and below the
quartiles) were removed.
Table 7.1 Nutrient and Salinity Category Benchmarks for NRSA Assessment
1 V'oivi'ion
S;ililiilv ;is
( OIldllClIN ll\
( uS am
( uuhJ-Liii'
S;ilinil> ;is
( undiiaiN ll\
(||S ail)
|-';iii'-l\uii'
Tnl;il \
MIU 1.)
(iiuid-Liir
Tnl;il \
m u 1.)
I';iii'-I\uii'
Tnl;il 1'
MIU 1.)
(iiuปd-l;;iir
Tnl;il 1'
IjlU 1.)
I;;iii"-I\uii'
CPL
500
1000
624
1081
55.9
103
NAP
500
1000
345
482
17.1
32.6
SAP
500
1000
240
456
14.8
24.4
UMW
500
1000
583
1024
36.3
49.9
TPL
1000
2000
700
1274
88.6
143
NPL
1000
2000
575
937
64.0
107
SPL
1000
2000
581
1069
55.8
127
WMT
500
1000
139
249
17.7
41.0
XER
500
1000
285
529
52.0
95.9
7.3 Signal to Noise
To examine within-year variability of water chemistry data, analysts used the revisit sites from the
Wadeable Streams Assessment, NRSA 2008-09, and NRSA 2013-14 to calculate signaknoise (S:N)
estimates for the national dataset. The results were a S:N ratio of 12.3 for total nitrogen, 10.2 for
total phosphorus, 31.2 for conductivity, and 39.2 for ANC.
7.4 Literature Cited
Baker, L.A., P.R. Kaufmann, A.T. Herlihy, and J.M. Eilers. 1990. Current Status of Surface Water
Acid-Base Chemistry. State of Science/Technology Report 9. National Acid Precipitation
Assessment Program, Washington, D.C., 650 pp.
Herlihy, A.T., and J.C. Sifneos. 2008. Developing nutrient criteria and classification schemes for
wadeable streams in the conterminous US. Journal of the North American Benthological Society
27:932-948.
101
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Kaufmann, P.R., A.T. Herlihy, M.E. Mitch, J.J. Messer, and W.S. Overton. 1991. Chemical
characteristics of streams in the Eastern United States: I. Synoptic survey design, acid-base
status and regional chemical patterns. Water'Resources Research 27:611-627.
USEPA. 2000. Nutrient Criteria Technical Guidance Manual for Rivers and Streams. EPA 822-B-
00-002. US. Environmental Protection Agency, Office of Water, Office of Science and
Technology, Washington, DC.
102
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
8 Physical Habitat Assessment
An assessment of river and stream (fluvial) physical habitat (PHab) condition is a major component
of the National Rivers and Streams Assessment (NRSA). Of many possible general and specific
fluvial habitat indicators measured in the NRSA surveys, the assessment team chose streambed
stability and excess fine sediments, instream habitat cover complexity, riparian vegetation, and
riparian human disturbances for the 2013-14 assessment. These four indicators have been used in
earlier U.S. Environmental Protection Agency (USEPA) national assessments, are important
nationwide, can be reliably and economically measured, and their reference conditions and degree
of anthropogenic alteration can be interpreted with reasonable confidence (Paulsen et al., 2008).
In the broadest sense, fluvial habitat includes all physical, chemical, and biological attributes that
influence or sustain organisms within streams or rivers. We use the term physical habitat to refer to
the structural attributes of habitat. NRSA made field measurements aimed at quantifying eight
general attributes of physical habitat condition, including direct measures of human disturbance.
Habitat Volume / Stream Siz e
Habitat Complexity and Cover for Aquatic Biota
Streambed Particle Size
Bed Stability and Hydraulic Conditions
Channel-Riparian and Floodplain Interaction
Hydrologic Regime
Riparian Vegetation Cover and Structure
Riparian Disturbance
These attributes were previously identified during EPA's 1992 national stream monitoring
workshop (Kaufmann 1993) as those essential for evaluating physical habitat in regional monitoring
and assessments. They are typically incorporated in some fashion in regional habitat survey
protocols (Platts et al. 1983, Fitzpatrick et al. 1998, Lazorchak et al. 1998, Peck et al. 2006, USEPA
2004) and were applied in the NRSA 2008-09 assessment (USEPA 2016), the National Wadeable
Streams Assessment (WSA: USEPA 2006) and the Western Rivers and Streams Pilot (EMAP-W)
surveys conducted between 2000 and 2005 (Stoddard et al. 2005a, b). The major habitat metrics
used in those past assessments and considered in NRSA are listed and defined in Table 8.1. Some
measures of these attributes are useful measures of habitat condition in their own right {e.g., channel
incision as a measure of channel-riparian interaction); others are important controls on ecological
processes and biota {e.g., bed substrate size), still others are important in the computation of more
complex habitat condition metrics {e.g., bankfull depth is used to calculate Relative Bed Stability
[RBJ]). Like biological characteristics, most habitat attributes vary according to their geomorphic
and ecological setting. Even direct measures of riparian human activities and disturbances are
strongly influenced by their geomorphic setting. And even within a region, differences in
precipitation and stream drainage area channel gradient (slope) lead to variation in many aspects of
103
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
stream habitat. Those geoclimatic factors influence discharge, flood stage, stream power (the
product of discharge times gradient), bed shear stress (proportional to the product of depth and
slope), and riparian vegetation. However, all eight of the major habitat attributes can be directly or
indirectly altered by anthropogenic activities.
NRSA follows the precedent of EMAP-W and WSA in reporting the condition of fluvial physical
habitat condition on the basis of four habitat indicators that are important nationwide, can be
reliably and economically measured, and their reference condition under minimal anthropogenic
disturbance can be interpreted with reasonable confidence. These are: relative bed stability (RBS) as
an indicator of bed sedimentation or hydrologic alteration, the areal cover and variety of fish
concealment features as a measure of in-stream habitat complexity, riparian vegetation cover and
structure as an indicator of riparian vegetation condition, and a proximity-weighted tally of
streamside human activities as an indicator of riparian human disturbances (Paulsen et al., 2008).
In this document, we describe the approach taken by NRSA in 2013-14 for assessing physical
habitat condition in rivers and streams based on the four above-mentioned indicators. We revisited
the screening of reference sites, consistently defining a set of reference sites from the combined
2013-14 and 2008-09 NRSA surveys, thereby increasing the number of sample sites available for
modeling expected condition, and for evaluating precision and responsiveness. We recalculated
PHab condition assignments in all previous surveys using the current NRSA 2013-14 assessment
procedures described here for our estimates of change or trends in PHab. We also examined the
rationale, importance, and measurement precision of each of the four indicators, including the
analytical approach for estimating reference conditions for each. Reference conditions for each
indicator were interpreted as their expected value in sites having the least amount of anthropogenic
disturbance within appropriately stratified regions. In most cases, we also refine the expected values
as a function of geoclimatic controlling factors within regions. Finally, we examine patterns of
association between physical habitat indicators and anthropogenic disturbance by contrasting
habitat indicator values in least-, moderately-, and most-disturbed sites nationally and within
regions.
Physical habitat metrics and condition assessment data from NRSA are available to download from
the NARS datawebpage - https://www.epa.gov/national-aquatic-resource-surveys/data-national-
aquatic-resource-surveys
8.1 Methods
8.1.1 Physical Habitat Sampling and Data Processing
Sample sites visited in NRSA are shown in Figure 8.1. In the wadeable streams sampled in NRSA,
field crews took measurements while wading the length of each sample reach (Peck et al. 2006); in
non-wadeable rivers, these measurements were made from boats (Hughes and Peck 2008). Physical
104
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
habitat data were collected from longitudinal profiles and from 11 cross-sectional transects and
streamside riparian plots evenly spaced along each sampled stream reach (USEPA 2007, 2013a, b).
The length of each sampling reach was defined proportional to the wetted channel width, and
measurements were placed systematically along that length to represent the entire reach. Sample
reach lengths were 40 times the wetted channel-width (ChW) long in wadeable streams and rivers,
with a minimum reach length of 150 m for channels less than 3.5 m wide. In non-wadeable
(boatable) rivers, reach lengths were also set to 40 ChW with a maximum length of 2,000 m.
Thalweg (maximum) depth measurements (in the deepest part of channel), habitat classification,
and mid-channel substrate observations were made at tightly spaced intervals; whereas channel
cross-sections and shoreline-riparian stations for measuring or observing substrate, fish cover
(concealment features), large woody debris, bank characteristics and riparian vegetation structure
were spaced further apart. Thalweg depth was measured at points evenly spaced every 0.4 ChW
along these reaches to give profiles consisting of 100 measurements (150 in streams <2.5 m wide).
The tightly spaced depth measures allow calculation of indices of channel structural complexity,
objective classification of channel units such as pools, and quantification of residual pool depth,
pool volume, and total stream volume. Channel slope and sinuosity on non-wadeable rivers were
estimated from l:24,000-scale digital topographic maps.
In wadeable streams and rivers, wetted width was measured and substrate size and embeddedness
were evaluated using a modified Wolman pebble count of 105 particles spaced systematically along
21 equally spaced cross-sections (Faustini and Kaufmann 2007), in which individual particles were
classified visually into seven size-classes plus bedrock, hardpan and other (e.g., organic material).
The numbers of pieces of large woody debris in the bankfull channel were tallied in 12 size classes
(3 length by 4 width classes) along the entire length of sample reaches. Channel incision and the
dimensions of the wetted and bankfull stream channel were measured at 11 equally-spaced
transects. Bank characteristics and areal cover of fish concealment features were visually assessed in
10 m long instream plots centered on transects, while riparian vegetation structure, presence of
large (legacy) riparian trees, non-native (alien) riparian plants, and evidence of human disturbances
(presence/absence and proximity) in 11 categories were visually assessed on adjacent 10 x 10 m
riparian plots on both banks. NRSA 2013-14 did not assess presence of large (legacy) trees and
non-native (alien) riparian plants, as had been done in previous surveys. In addition, channel
gradient (slope) in wadeable streams was measured to provide information necessary for calculating
residual pool depth and RBS. In wadeable streams, crews used laser or hydrostatic levels, but if
necessary, were allowed to use hand-held clinometers in channels with slopes >2.5%. Compass
bearings between stations were obtained for calculating channel sinuosity. Channel constraint and
evidence of debris torrents and major floods were assessed over the whole reach after the other
components were completed. Discharge was measured by the velocity-area method at the time of
sampling, or by other approximations if that method was not practicable (Peck et al. 2006; USEPA
2007, 2013a, b).
In boatable rivers, NRSA field crews measured the longitudinal thalweg depth profile
105
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
(approximated at mid-channel) using 7.5 m telescoping survey rods or SONAR. At the same time,
crews tallied snags and off-channel habitats, classified main channel habitat types, and characterized
mid-channel substrate by probing the bottom. At 11 littoral/riparian plots (each 10 m wide x 20 m
long) spaced systematically and alternating sides along the river sample reach, field crews measured
channel wetted width, bankfull channel dimensions, incision, channel constraint. They assessed
near-shore, shoreline, and riparian physical habitat characteristics by measuring or observing littoral
depths, riparian canopy cover, substrate, large woody debris, fish cover, bank characteristics,
riparian vegetation structure, presence of large ("legacy") riparian trees, non-native riparian plants,
and evidence of human activities. As was the case for wadeable streams, NRSA 2013-14 did not
assess presence of large (legacy) trees and non-native (alien) riparian plants in boatable rivers, as had
been done in previous surveys. After all the thalweg and littoral/riparian measurements and
observations were completed, the crews estimated the extent and type of channel constraint (see
USEPA 2007, 2013a, b). Channel slope and sinuosity on all boatable rivers were estimated from
l:24,000-scale digital topographic maps.
See Kaufmann et al. (1999) for calculations of reach-scale summary metrics from field data,
including mean channel dimensions, residual pool depth, bed particle size distribution, wood
volume, riparian vegetation cover and complexity, and proximity-weighted indices of riparian
human disturbances. See Faustini and Kaufmann (2007) for details on the calculation of geometric
mean streambed particle diameter, Kaufmann et al. (2008, 2009) for calculation of bed shear stress
and relative bed stability (modified since published by Kaufmann et al. 1999), and Kaufmann and
Faustini (2012) for demonstrating the utility of EMAP and NRSA channel morphologic data to
estimate transient storage and hydraulic retention in wadeable streams.
8.1.2 Quantifying the Precision of Physical Habitat Indicators
The absolute and relative precision of the physical habitat condition metrics used in NRSA are
shown in Table 8.2, based on data from 4,193 sites (2,113 from NRSA 2008-09 and 2,080 from
NRSA 2013-14) and repeat visits to a random subset of 388 of those sites (197 and 191 revisits in
the two surveys). The RMSrep expresses the precision or replicability of field measurements,
quantifying the average variation in a measured value between same-season site revisits, pooled
across all sites where measurements were repeated. We calculated RMSrep as the root-mean-square
error of repeat visits during the same year, equivalent to the pooled standard deviation of repeat
visits relative to their site means, as discussed Kaufmann et al. (1999) and Stoddard et al. (2005a).
S:N is the ratio of variance among streams ("signal") to that for repeat visits to the same stream
("noise") as described by Kaufmann et al. (1999).
The ability of a monitoring program to detect trends is sensitive to the spatial and temporal
variation in the target indicators as well as the design choices for the network of sites and the
timing and frequency of sampling. Sufficient temporal sampling of sites was not available to
106
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
estimate all relevant components of variance for the entire U.S. However, Larsen et al. (2004)
examined the survey sampling variance components for a number of the EMAP-NARS physical
habitat variables, including some of interest in this chapter (residual depth, canopy cover, fine
sediment, and in-channel large wood). Their analysis was based on evaluation on six Pacific
Northwest surveys that included 392 stream reaches and 200 repeat visits. These surveys were
conducted in Oregon and Washington from 1993 to 1999. Most were from one to three years in
duration, but one survey lasted six years. They modeled the likelihood of detecting a 12% per year
trend in the selected physical habitat characteristics, if such a trend occurs, as a function of the
duration of a survey. To calculate the number of years required to detect the defined trends in a
monitoring network with a set number of sites, they set the detection probability at >80% with
<5% probability of incorrectly asserting a trend if one is not present. We used the same survey data
sets to duplicate their analysis for several variables not included in the Larsen et al. (2004)
publication, including log transformed relative bed stability (LRBS_BWS) and riparian vegetation
cover complexity (XCMGW, the combined cover of three layers of riparian woody vegetation); the
results of that trend detection potential is summarized in Table 8.3.
8.2 Physical Habitat Condition Indicators
8.2.1 Relative Bed Stability and Excess Fines
Streambed characteristics (e.g., bedrock, cobbles, silt) are often cited as major controls on the
species composition of macroinvertebrate, periphyton, and fish assemblages in streams (e.g., Hynes
1970, Cummins 1974, Platts et al. 1983, Barbour et al. 1999, Bryce et al. 2008, 2010). Along with
bedform (e.g., riffles and pools), streambed particle size influences the hydraulic roughness and
consequently the range of water velocities in a stream channel. It also influences the size range of
interstices that provide living space and cover for macroinvertebrates and smaller vertebrates.
Accumulations of fine substrate particles (excess fine sediments) fill the interstices of coarser bed
materials, reducing habitat space and its availability for benthic fish and macroinvertebrates
(Hawkins et al. 1983, Platts et al. 1983, Rinne 1988). In addition, these fine particles impede
circulation of oxygenated water into hyporheic habitats reducing egg-to-emergence survival and
growth of juvenile salmonids (Suttle et al. 2004). Streambed characteristics are often sensitive
indicators of the effects of human activities on streams (MacDonald et al. 1991, Barbour et al. 1999,
Kaufmann et al. 2009). Decreases in the mean particle size and increases in streambed fine
sediments can destabilize stream channels (Wilcock 1997, 1998) and may indicate increases in the
rates of upland erosion and sediment supply (Lisle 1982, Dietrich et al. 1989).
"Unsealed" measures of surficial streambed particle size, such as percent fines or Dso, can be useful
descriptors of streambed conditions. In a given stream, increases in percent fines or decreases in
Dfomay result from anthropogenic increases in bank and hillslope erosion. However, a great deal of
the variation in bed particle size among streams is natural: the result of differences in stream or
river size, slope, and basin lithology. The power of streams to transport progressively larger
107
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
sediment particles increases in direct proportion to the product of flow depth and slope. All else
being equal, steep streams tend to have coarser beds than similar size streams on gentle slopes.
Similarly, the larger of two streams flowing at the same slope will tend to have coarser bed material,
because its deeper flow has more power to scour and transport fine particles downstream (Leopold
et al. 1964, Morisawa 1968). For these reasons, we "scale" bed particle size metrics, expressing bed
particle size in each stream as a deviation from that expected as a result of its size, power, and
landscape setting (Kaufmann et al., 1999, 2008, 2009).
The scaled median streambed particle size is expressed as Relative Bed Stability (RBS), calculated as
the ratio of the geometric mean diameter, Dm, divided by Dcm, the critical diameter (maximum
mobile diameter) at bankfull flow (Gordon et al, 1992), where Dgm is based on systematic
streambed particle sampling ("pebble counts") and Dcm is based on the estimated streambed shear
stress calculated from slope, channel dimensions, and hydraulic roughness during bankfull flow
conditions.
RBS is a measure of habitat stability for aquatic organisms as well as an indication of the potential
for economic risk to streamside property and structures from stream channel movement. In many
regions of the U.S., we may also be able to use RBS to infer whether sediment supply is augmented
by upslope or bank erosion from anthropogenic or other disturbances, because it can indicate the
degree of departure from a balance between sediment supply and transport. In interpreting RBS on
a regional scale, Kaufmann et al (1999, 2009) argued that, over time, streams and rivers adjust
sediment transport to match supply from natural weathering and delivery mechanisms driven by
the natural disturbance regime, so that RBS in appropriately stratified regional reference sites
should tend towards a range characteristic of the climate, lithology, and natural disturbance regime.
Values of the RBS index either substantially lower (finer, more unstable streambeds) or higher
(coarser, more stable streambeds) than those expected based on the range found in least-disturbed
reference sites within an ecoregion are considered to be indicators of ecological stress.
Excess fine sediments can destabilize streambeds when the supply of sediments from the landscape
exceeds the ability of the stream to move them downstream. This imbalance can result from
numerous human uses of the landscape, including agriculture, road building, construction, and
grazing. Lower-than-expected streambed stability may result either from high inputs of fine
sediments (from erosion) or increases in flood magnitude or frequency (hydrologic alteration).
When low RBS results from fine sediment inputs, stressful ecological conditions result from fine
sediments filling in the habitat spaces between stream cobbles and boulders (Bryce et al 2008,
2010). Instability (low RBS) resulting from hydrologic alteration can be a precursor to channel
incision and arroyo formation (Kaufmann et al 2009). Perhaps less well recognized, streams that
have higher than expected streambed stability can also be considered stressedvery high bed
stability is typified by hard, armored streambeds, such as those often found below dams where fine
sediment flows are interrupted, or within channels where banks are highly altered. Values of RBS
higher than reference expectations can indicate anthropogenic coarsening or armoring of
108
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
streambeds, but streams containing substantial amounts of bedrock may also have very high RBS,
and at this time it is difficult to determine the role of human alteration in stream coarsening on a
national scale. For this reason, NRSA reported only on the "low end" of RBS relative to reference
conditions, generally indicating streambed excess fine sediments or augmented stormflows
associated with human disturbance of stream drainages and riparian zones.
8.2.1.1 Precision of Sediment and Bed Stability Measurements
The geometric mean bed particle diameter (Dgm) and RBS, respectively, varied over 6 and 9 orders
of magnitude in the NRSA surveys. Because of this wide variation and the fact that both exhibit
repeat-visit variation that is proportional to their magnitude at individual streams, it is useful and
necessary to log transform these variables (LSUB_DMM and LRBS^08). The RMSrep of
LSUB_DMM in the two combined NRSA surveys was 0.39, but the wadeable stream "pebble
count" procedure was more precise (RMSrep=0.25) than the bottom-probing procedure applied in
boatable rivers (RMSrep=0.51). For NRSA's wadeable streams, this precision for LSUB-DMM was
similar to that reported by Faustini and Kaufmann (2007) for EMAP-W (0.21). For a Dm = "y"
mm, the log-based RMSrep of 0.25 translates to an asymmetrical 1SD error bound of 0.56>< to 1 .ISj
mm, and for a log-based RMSrep of 0.51, a 1SD error bound of 0.31j> to 3.24j mm.
The RMSreP of LRBS^08 in NRSA wadeable and boatable sites was 0.52, approximately 6% of its
observed range, but less precise (surprisingly) than that for EMAP-W (RMSrep = 0.365). The log-
based RMSrep of 0.52 for NRSA LRBS_g08 translates to an asymmetrical error bound of 0.30j to
3.3j around an untransformed RBS value of "y" (Table 8.2). Compared with the high S:N ratio for
LSUB_DMM in NRSA wadeable+boatable waters (S:N=10.9), relative precision for LRBS^g08 was
lower (S:N=4.2), reflecting the reduction in total variance that results from "modeled out" a large
component of natural variability by scaling for channel gradient, water depth, and channel
roughness. Nevertheless, the moderate relative precision of LRBS^g08 is easily adequate to make it
a useful variable in regional and national assessments (Kaufmann et al. 1999, 2008, Faustini and
Kaufmann 2007). The transformation of the unsealed geometric mean bed particle diameter Dgm to
the ratio RJ5S by dividing by the critical diameter reduced the within-region variation by accounting
for some natural controlling factors. As a result, we feel that the scaled variable helps to reveal
alteration of bed particle size and mobility from anthropogenic erosion and sedimentation
(Kaufmann et al. 2008, 2009).
We have examined the components of variability of LRBS based on earlier surveys and modeled its
potential utility in trend detection in the Pacific Northwest region of the U.S. with the same data
and procedures as used by Larsen et al. (2004), in which all methods were the same as used in
EMAP-W and WSA except that bed substrate mean diameter data used by Larsen et al. was
determined based on 55, rather than 105 particles. (NRSA data differed from data used in that
analysis by using laser levels rather than hand-held clinometers to measure wadeable stream slopes
<2.5%) That analysis showed that a 50-site monitoring program could detect a subtle trend in
109
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
LIIBS_BWS of 2% per year within 8 years, if sites were visited every year (Table 8.3).
8.2.2 Riparian Vegetation
8.2.2.1 Quantifying Riparian Vegetation Cover Complexity
The importance of riparian vegetation to channel structure, cover, shading, inputs of nutrients and
large wood, and as a wildlife corridor and buffer against anthropogenic disturbance is well
recognized (Naiman et al. 1988, Gregory et al. 1991). Riparian vegetation not only moderates stream
temperatures through shading, but also increases bank stability and the potential for inputs of
coarse and fine particulate organic material. Organic inputs from riparian vegetation become food
for stream organisms and provide structure that creates and maintains complex channel habitat.
The presence of a complex, multi-layered vegetation corridor along streams and rivers is an
indicator of how well the stream network is buffered against sources of stress in the watershed.
Intact riparian areas can help reduce nutrient and sediment runoff from the surrounding landscape,
prevent streambank erosion, provide shade to reduce water temperature, and provide leaf litter and
large wood that serve as food and habitat for stream organisms (Gregory et al., 1991). The presence
of large, mature canopy trees in the riparian corridor reflects its longevity, whereas the presence of
smaller woody vegetation typically indicates that riparian vegetation is reproducing, and suggests
the potential for future sustainability of the riparian corridor (Kaufmann and Hughes 2006).
NRSA evaluated the cover and complexity of riparian vegetation based on the metric XCMGW,
which is calculated from visual estimates made by field crews of the areal cover and type of
vegetation in three layers: the ground layer (<0.5 m), mid-layer (0.5-5.0 m) and upper layer (>5.0
m). The separate measures of large and small diameter trees, woody and non-woody mid-layer
vegetation, and woody and non-woody ground cover are all visual estimates of areal cover.
XCMGW sums the cover of woody vegetation over these three vegetation layers, expressing both the
abundance of vegetation cover and its structural complexity. Its theoretical maximum is 3.0 if there
is 100% cover in each of the three vegetation layers. XCMGW gives an indication of the longevity
and sustainability of perennial vegetation in the riparian corridor (Kaufmann et al. 1999, Kaufmann
and Hughes 2006).
8.2.2.2 Precision of Riparian Vegetation Index
XCMGW ranged from 0 to 2.6 (260% cover), with RMSrep of Log(0.01+XCMGW) = 0.148 (Table
8.2), meaning that an XCMGW value of 10% at a single stream site has a ฑ1.0 RMSrep error bound
of 7% to 14%. Its S:N ratio was 8.45, indicating very good potential for discerning differences
among sites. We examined the components of variability of XCMGW and modeled its potential
utility in trend detection in the Pacific Northwest region of the U.S. with the same data and
procedures used by Larsen et al. (2004). Based on that analysis, a 50-site monitoring program could
detect a subtle trend in XCMGW of 2% per year within 8 years, if sites were visited every year
110
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
(Table 8.3).
8.2.3 Instream Habitat Cover Complexity
Although the precise mechanisms are not completely understood, the most diverse fish and
macroinvertebrate assemblages are usually found in streams that have complex mixtures of habitat
features: large wood, boulders, undercut banks, tree roots, etc. (see Kovalenko et al. 2011). When
other needs are met, complex habitat with abundant cover should generally support greater
biodiversity than simple habitats that lack cover (Gorman and Karr 1978, Benson and Magnuson
1992). Human use of streams and riparian areas often results in the simplification of this habitat,
with potential effects on biotic integrity (Kovalenko et al. 2011). For this assessment, we use a
measure (XFC_NAT in Kaufmann et al. 1999) that sums the amount of instream habitat consisting
of undercut banks, boulders, large pieces of wood, brush, and cover from overhanging vegetation
within a meter of the water surface, all of which were estimated visually by NRSA field crews.
8.2.3.1 Quantifying Instream Habitat Complexity
Habitat complexity is difficult to quantify, and could be quantified or approximated by a wide
variety of measures. The NRSA Physical Habitat protocols provide estimates for nearly all of the
following components of complexity identified during EPA's 1992 stream monitoring workshop
(Kaufmann 1993):
Habitat type and distribution {e.g., Bisson et al. 1982, O'Neill and Abrahams 1984, Frissell et
al. 1986, Hankin and Reeves 1988, Hawkins et al. 1993, Montgomery and Buffington 1993,
1997, 1998).
Large wood count and size (e.g., Harmon et al. 1986, Robison and Beschta 1989, Peck et al.
2006).
In-channel cover: percentage areal cover of fish concealment features, including undercut
banks, overhanging vegetation, large wood, boulders (Hankin and Reeves 1988, Kaufmann
and Whittier 1997, Peck et al. 2006).
Residual pools, channel complexity, hydraulic roughness (e.g., Kaufmann 1987a, b, Lisle
1987, Stack and Beschta, 1989; Lisle and Hilton 1992, Robison and Kaufmann 1994,
Kaufmann et al. 1999, Kaufmann et al. 2008, Kiem et al, 2002; Kaufmann et al. 2011).
Width and depth variance, bank sinuosity (Kaufmann 1987a, Moore and Gregory 1988,
Kaufmann et al. 1999, Madej 1999, 2001, Kaufmann et al. 2008, Mossop and Bradford 2006,
Pearsons and Temple 2007, 2010, Kaufmann and Faustini 2012).
Residual depth is a measure of habitat volume, but also serves as one of the indicators of channel
habitat complexity, particularly when expressed as a deviation from reference expectations,
including the influences of basin size. A stream with more complex bottom profile will have greater
residual depth than one with similar drainage area, discharge and slope, but lacking that complexity
111
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
(Kaufmann 1987a). Conversely, between two streams of equal discharge and slope, the one with
greater residual depth (i.e., larger, more abundant residual pools) will have greater variation in cross-
sectional area, slope, and substrate size. A related measure of the complexity of channel
morphology is the coefficient of variation in thalweg depth, calculated entirely from the thalweg
depth profile (SDDEPTH / XDEPTH). The thalweg profile is a systematic survey of depth in the
stream channel along the path of maximum depth (i.e., the thalweg). In addition to measures of
channel morphometric complexity, NRSA physical habitat protocols measure in-channel large
wood (sometimes called "large woody debris" or simply "LWD"), and several estimates of the areal
cover of various types of fish and macroinvertebrate "cover" or concealment features. The large
wood metrics include counts of wood pieces per 100 m of bankfull channel and estimates of large
wood volume in the sample reach expressed in cubic meters of wood per square meter of bankfull
channel. The "fish cover" variables are visual estimates of the areal cover of single or combined
types of habitat features.
NRSA required a general summary metric as a holistic indicator of many aspects of habitat
complexity, so NRSA used the metric XFC_NsiT, summing the areal cover from large wood,
brush, overhanging vegetation, live trees and roots, boulders, rock ledges, and undercut banks in
the wetted stream channel. Habitat complexity and the abundance of particular types of habitat
features differ naturally with stream size, slope, lithology, flow regime, and potential natural
vegetation. For example, boulder cover will not occur naturally in streams draining deep deposits of
loess or alluvium that do not contain large rocks. Similarly, large wood will not be found naturally
in streams located in regions where riparian or upland trees do not grow naturally. Though the
index XFC_NslT partially overcomes these differences by summing divergent types of cover, we
set stream-specific expectations for habitat complexity metrics in NRSA based on region-specific
reference sites and further refined them as a function of geoclimatic controls.
8.2.3.2 Precision of habitat complexity measures
The instream habitat complexity index XFC_N,AT ranged from 0 to 2.3, or 0% to 230% in NRSA
(2008-09 and 2013-14 combined), expressing the combined areal cover of the five cover elements
contributing to its sum. The RMSrep of Log(0.01+XFC_AM7} was 0.21, meaning that an
XFC_Ny4T value of 10% cover at a single stream site has a +1.0 RMSrep error bound of 6% to 16%
(Table 8.2). S:N was relatively low for this indicator (2.27), though higher in wadeable streams
(2.76) than in boatable rivers (1.66). Despite its relatively low S:N, the RMSrep for LXFC_Ny4Twas
9% of its observed range. It was retained as a habitat complexity indicator because it contains
biologically relevant information not available in other metrics, showed moderate responsiveness to
human disturbances, and has precision adequate to discern relatively large differences in habitat
complexity.
112
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
8.2.4 Riparian Human Disturbances
Agriculture, roads, buildings, and other evidence of human activities in or near stream and river
channels may exert stress on aquatic ecosystems and may also serve as indicators of overall
anthropogenic stress. EPA's 1992 stream monitoring workshop recommended field assessment of
the frequency and extent of both in-channel and near-channel human activities and disturbances
(Kaufmann 1993). The vulnerability of the stream network to potentially detrimental human
activities increases with the proximity of those activities to the streams themselves. NRSA follows
Stoddard et al. (2005b) and U.S. EPA (2006) in using a direct measure of riparian human
disturbance that tallies 11 specific forms of human activities and disturbances (walls, dikes,
revetments or dams; buildings; pavement or cleared lots; roads or railroads; influent or effluent
pipes; landfills or trash; parks or lawns; row crop agriculture; pasture or rangeland; logging; and
mining) at 22 separate locations along the stream reach, and weights them according to how close
to the channel they are observed (W1 _HAJ J, in Kaufmann et al. 1999). Observations within the
stream or on its banks are weighted by 1.5, those within the 10 x 10 m plots are weighted by 1.0,
and those visible beyond the plots are weighted by 0.5. The index W1 _HA IJ, ranged from 0 (no
observed disturbance) to ~1 {e.g., equivalent to four or 5 types of disturbance observed in the
stream, throughout the reach; or seven types observed within all 22 riparian plots bounding the
stream reach). Although direct human activities certainly affect riparian vegetation complexity and
layering measured by the Riparian Vegetation Index (previous paragraph), the Riparian Disturbance
Index is more encompassing, and differs by being a direct measure of observable human activities
that are presently or potentially detrimental to streams.
8.2.4.1 Precision of riparian disturbance indicators
The proximity-weighted human disturbance indicator W1 _HA1 J, ranged from 0 to 8.3 in NRSA,
and its precision was proportional to the level of disturbance. The RMSrep of log(0.l + W1 _HAJ J )
was 0.178 (Table 8.2), meaning that a W1 _HAJ J, value of 1.0 at a single stream site has a +1.0
RMSrep error bound of 0.66 to 1.51. The relative precision of Log(0.1 + W1 _HAJ J,) was moderate
(S:N=5.46), indicating good potential for discerning differences among sites.
8.3 Estimating Reference Condition for Physical Habitat
8.3.1 Reference Site Screening and Anthropogenic Disturbance Classifications
As part of the routine application of its field and GIS protocols, NRSA (2008-09 and 2013-14
combined) obtained various measures of human disturbance associated with each site and its
catchment. Site-scale indicators of human disturbance included field observations of various human
activities including nearby roads, riprap, agricultural activities, riparian vegetation disturbance, etc.,
as detailed by Kaufmann et al. (1999). These indicators of local scale disturbance were used in
combination with water chemistry (Chloride, Total Phosphorus, Total Nitrogen, Sulfate, and
Turbidity), as described in Section 4 and by Herlihy et al. (2008), to screen probability and hand-
113
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
picked sites and designate them as least- moderately-, and most-disturbed, relative to other sites
within each of the regions of NRSA. To avoid circularity, we did not use any field measures of
sediment, in-channel habitat complexity, or riparian vegetation to screen least-disturbed sites used
to estimate reference condition for excess streambed fining, instream fish cover, and riparian
vegetation. Nor did we use such measures in defining levels of disturbance to use in examining the
associations of these habitat metrics with human disturbances. We did, however, use field
observations of the level and proximity of streamside human activities (W1-HAJ J,, W1 _HAG,
W1H_CROP, and W1 U_\VA T J,) in screening reference sites and defining levels of disturbance for
evaluating indicator responsiveness (Table 8.4). In this chapter, the designation "R" refers to least-
disturbed ("reference") sites; "S" to moderately-disturbed sites, and "T" to the most-disturbed sites
within each of the nine aggregate ecoregions discussed herein. We defined these site disturbance
categories independent of the habitat indicators we evaluate in this report (other than riparian
human disturbances), allowing an assessment of fluvial habitat response to a gradient of human
activities and disturbances. We also used sub-basin row crop and urban land use percentages, and
the density of dams and impoundments to reject potential reference sites.
Screening the NRSA 2008-09 and 2013-14 survey sites by the disturbance variables in Table 8.4
yielded 708 reference sites, 349 from the first survey and 359 from the second (Table 8.5). Fewer
reference sites were identified for boatable (281) than for wadeable (427) streams and rivers; except
in several regions, reference sites were approximately evenly distributed between the two surveys.
Notably, only 2 boatable reference sites were identified in the SPL (both in the 2008-09 survey),
and only 7 boatable and 7 wadeable reference sites were identified in the SAP. Interestingly, more
reference sites were identified in the 2013-14 survey of the CENPL and Western regions than in
the 2008-09 survey of those regions. The opposite was true of the Appalachians.
8.3.2 Modeling Expected Reference Values of the Indicators
8.3.2.1 Modeling Approaches
In the following paragraphs, we describe the conceptual basis for modeling the expected range of
values for the each of the physical habitat indicators under least-disturbed (reference) condition.
The details of these models are presented in Table 8.6 Table 8.8, and with more detail in
Appendix 8.A. For riparian human activities, we applied uniform criteria based on professional
judgement and literature to assign high, medium and low disturbance to individual sample sites
across the entire U.S. For the other three PHab indicators, we assigned habitat condition based on
the distribution of PHab metric values within the combined set of NRSA reference sites, employing
several types of modeling:
NULL MODELS based expected least-disturbance values and their distribution on the mean and
SD of the indicator metric (e.g., LRBS^g08, XCMGW, or XFC_Nai) in the set of reference sites
representing least-disturbed condition within resource types (e.g., wadeable and boatable) in their
respective regions (ECOmaP) or aggregations of those regions (e.g., Central Plains = CENPL =
114
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
NPL+SPL+TPL). For example, in NAP boatable sites, LRBS^g08 null model condition classes
were defined based on normal approximations of the 5th and 25th percentiles of the actual reference
distributions. The definition of "Poor" condition was set for those sites with LRBS^g08 < the
reference mean LKBS^08 minus 1.65(SDref). Sites in "Good" condition with respect to this
indicator were those with LRBS^g08 > the reference mean LRBS^g08 minus 0.67(SDref). As for
RBS_g08, we log-transformed XCMGW and XFC_Nat to approximate statistical normality in
distributions (e.g., LRBS^g08 = LogiofRBi^O#], L,Pt01 _XCMGW = Logio[0.01 + XCMGW], and
LPt01 _XFC_Nat = Logio[0.01+XFC_I\to]).
REFERENCE-SITE OBSERVED/EXPECTED (O/E) MODELS: In cases where reference
sites were sufficiently numerous and spanned a representative range of the natural controlling
variables, we applied Multiple Linear Regression (MLR) to regional reference sites (only) in order to
factor out the influence of natural controlling factors on habitat separate from the influences of
anthropogenic disturbances. These MLR models estimate site-specific expected values of habitat
metrics under least-disturbed conditions, given their geoclimatic and geomorphic setting (e.g.,
ecoregion, latitude, longitude, drainage area, channel width, slope, elevation, and soil erodibility). If
there were less than 22 reference sites in a region, or we determined that reference sites may not
fully encompass the geoclimatic variables controlling a habitat metric, we combined regions with
similar controlling factors in the modelling. The variables made available to MLR were
LAT_DD83, LON_DD83, L_AreaWSkm2_use, ELEV_PT_use, UXSlope_use, UXWidth_use, and
KFCT_WS_use. We then calculated observed/expected (O/E) values of the habitat metrics for
every site within the modelled region, including non-reference sites. We set expectations of the
O/E values based on the mean and SD of the O/E values in the regional reference distribution,
and set Good, Fair, and Poor condition determinations based on normal approximation of log-
transformed O/E values as described for the LKBS null model in the previous paragraph.
REFERENCE-SITE O/E MODELS WITH DISTURBANCE ADJUSTMENT: In cases
where reference sites were sufficiently numerous and spanned a representative range of the natural
controlling variables, but had substantial anthropogenic disturbances that influenced the habitat
metric response variable, we included riparian and basin disturbance variable(s) as predictors in the
Reference Site MLRs. As with the Reference Site models with no adjustment, we combined regions
with similar controlling factors in the modelling, where the number and representativeness of
reference sites were inadequate in a given region. Besides the geoclimatic and geomorphic variables
listed in the previous paragraph, we considered the following disturbance variables in these MLRs:
W1_Hall, W1_HNOAG, W1_HAG, W1H_Crop, DAM_du, AGJKMCircle, URB_1 KMCircle,
RDDEN_WS_use, PCT_AG_WS_use, and AGm_X_KFct (interaction of basin % crop agriculture
with soil erodibility factor). Site-specific expected ("E") values of the habitat metric were then
calculated by setting the anthropogenic disturbance metric values to the lowest value observed
("O") among reference sites in the modelled region. Because we had already modeled-out
disturbance to some extent in our calculation of E values, the distributions of O/E in reference
sites did not necessarily have a mean of 1/1 (Log=0), although means were very close to 1/1. We
115
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
then calculated O/E values of the habitat metrics for every site within the modelled region,
including non-reference sites. We set expectations of the O/E values based on the mean and SD of
the distributions of Logio(0/E) values in regional reference sites, analogous to that described for
reference site regressions in the previous paragraph, and set Good, Fair, and Poor condition
determinations based on normal approximation of log-transformed O/E values analogous to that
described for the LRBS null model above.
"All-Sites" O/E MODELS: In cases where reference sites were generally disturbed and where the
number and distribution of minimally-disturbed reference sites were insufficient to accurately
quantify geoclimatic influences on a given habitat metric, we employed "All-Sites" O/E models.
We used two steps to calculate reference expected values. The first step was to calculate expected
values from MLRs that employed all sites (not just reference sites) in the model region, and
considered both geoclimatic and anthropogenic predictors. Site-specific expected ("E") values of
the habitat metric were then calculated using the MLR equation with the anthropogenic disturbance
metric values set to their lowest value observed ("O") in the modelled region. We then calculated
O/E values of the habitat metrics for every site within the modelled region. In the second step, we
examined the distribution of O/E values in reference sites and their association with anthropogenic
disturbance within the region. In cases where reference site O/E values showed no association with
disturbance, we based reference expectations on the mean and SD of the distributions of
Log 10(O/E) values in these regional reference sites, analogous to that described for unadjusted
regression site regressions in the previous paragraph. We then set Good, Fair, and Poor condition
determinations based on normal approximation of log-transformed O/E values analogous to that
described for the LRBS null model in the previous paragraph. In cases were reference site O/E
values were still associated with anthropogenic disturbance, our second step included regressing the
Log 10(O/E) values against anthropogenic disturbance variables to determine expected O/E values
under least-disturbed conditions. We then set the anthropogenic disturbance variables in the MLR
to their regional minimum values, effectively choosing the y-intercept of these equations as the
central tendency for expected reference condition. We set expectations of the O/E values based on
the y-intercept and regression RMSE of Logio(0/E) values in regional reference sites, analogous to
that described for unadjusted reference site regressions in the previous paragraph, and set Good,
Fair, and Poor condition determinations based on normal approximation of log-transformed O/E
values analogous to that described for the LRBS null model above.
8.3.2.2 Bed Sediment Condition Modeling
We used reference site null models to estimate expected reference values of Logio Relative Bed
Stability (LRBS_g08) in boatable rivers and streams in 5 of the 9 ecoregions (NAP, SAP, CPL,
WMT, and XER). RMSE's for these null models ranged from 0.365 in the WMT to 1.539 in the
NAP. Modeling for boatable sites in the 4 remaining regions were MLR models with R2 ranging
from 18% to 56%, RMSE from 0.365 to 1.539, and included one to three predictors. Predictors
were primarily drainage area (LAm), channel width (UXWidth), and extent of agricultural land use
116
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
in the contributing drainages, or within a 1 km radius of the sample sites on these rivers (Table 8.6,
and more detail in Appendix 8.A). For boatable rivers in the NPL, SPL, and TPL, we employed
All-Sites MLR models that incorporated similar predictors as those used in the reference site MLRs.
For wadeable streams in all except the Central Plains regions (NPL, SPL, and TPL), we used
reference site MLRs to estimate LoglO Relative Bed Stability (LRBS^08) in least-disturbed sites.
These MLRs most commonly included a basin or stream size variable (LAm or UXWidtfi), slope
('LXSlope), and usually a site-scale or basin measure of human land use intensity. In the NPL, SPL,
and TPL, we employed All-Sites MLR models typically incorporating Lat and/or Lon with LAm,
Elevation or Slope and one or more variables representing the intensity of human land use activity
in the drainage basin, vicinity, or near the banks of the sample reaches. MLR model R2 values
ranged from 20% to 41%, and RMSE ranged from 0.430 to 0.990. The reference site models had 1
to 3 predictors and the All-Sites models had 4 to 5 predictors.
8.3.2.3 Riparian Vegetation Cover & Structure Condition Modeling
Reference site null models were employed for estimating expected reference condition for Riparian
Vegetation Cover & Structure (LPt01 _XCMGW) only for boatable rivers in the TPL and WMT
(Table 8.7 with greater detail in Appendix 8.A). All-Sites MLR models were used for boatable
rivers in the combined NPL and SPL and for wadeable streams in the NPL. The boatable All-Sites
MLR incorporated Lat, Lon, site-level agriculture (W1 _HAG), basin road density (RDDEN_m),
and % of agricultural land use in the drainage basin (PCT_AG_m). The NPL wadeable stream All-
Sites model was similar, incorporating Lat, Lon, LXSlope, LXWidth, site-level agriculture
(W1 _HAG), basin road density (RDDEN_m), and PCT_AG_m. Expected condition models for
boatable or wadeable streams in all the remaining ecoregions were reference site regression models
with 1 to 4 geoclimatic predictors including Lat or Lon, along with LAm, LXWidth, LXSlope, or
Elev. Most of these MLRs also included one or more variables representing the intensity of human
land use activity in the drainage basin, vicinity, or near the banks of the sample reaches. Model R2
was 1% for CPL wadeable streams, and 14% to 40% elsewhere. The precision of these reference
site MLRs and All-Sites models (RMSE 0.119 to 0.487) was generally greater (smaller RMSE) for
these riparian vegetation models than for the LRBS models.
8.3.2.4 Instream Habitat Cover & Complexity Condition Modeling
Reference site null models were employed for estimating expected reference condition for Instream
Habitat Cover Complexity (LPt01 _XFC_Nat) only in the CPL, where we used separate null models
for wadeable and boatable sites (Table 8.8 with greater detail in Appendix 8.A). All the remaining
expected condition models were reference site regression models incorporating 1 to 5 predictors,
with R2 ranging from 7% to 53% and RMSE's from 0.175-0.335, somewhat less precise (larger)
than those for riparian vegetation condition. These expected condition MLRs typically included 1
to 3 predictors from the set of geoclimatic variables including Lat, Lon, LAm, LXWidth, LXSlope,
or Elev. Except for NAP and UMW wadeable stream MLRs and the XER boatable river model, all
117
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
the other instream habitat condition MLRs also included one or more variables representing the
intensity of human land use activity in the drainage basin, vicinity, or near the banks of the sample
reaches.
8.3.2.5 Riparian Human Disturbance Indicator Condition Determination
For the riparian human disturbance indicator, we did not base condition benchmarks on the
reference distributions or expected condition MLRs, as was done for bed sediments, riparian
vegetation condition and habitat complexity. Instead, we set these classes using uniform judgement-
based criteria for all regions. W1_Hall, the database variable name for this indicator, is a direct
measure of human disturbance "pressure," unlike the other habitat indicators, which are actually
measures of habitat response to human disturbance pressures. It is very difficult to define reference
sites without screening sites based on W1_Hall. For this reason, we took this different approach for
setting riparian disturbance benchmarks, defining low disturbance sites as those with W1 _Hall
<0.33 and high riparian disturbance sites as those with W1_Hซ//>1.5; we applied these same
benchmarks in all ecoregions. A value of 1.5 for a stream means, for example, that at 22 locations
along the stream the field crews found an average of one of 11 types of human disturbance within
the stream or its immediate banks. A value of 0.33 means that, on average, one type of human
disturbance was observed at one-third of the 22 riparian plots along a sample stream or river.
8.4 Response of the Physical Habitat Indicators to Human Disturbance
Riparian human disturbance (W1_HaII) values between 0 and 3 were found in all regions and in
both boatable and wadeable sites (Figure 8.2). Among regional reference sites, UMW boatable and
wadeable reference sites and WMT wadeable reference sites had the lowest riparian disturbance
(Figure 8.3). Very high values of W1_Hallwere found in all regions with the exception of
wadeable streams in the UMW (note tradition of riparian buffer protection that is visible from the
air), and steep gradients of W1_Hallwere found across the three disturbance classes in all regions
(Figure 8.3). Because the field-obtained measures of riparian disturbance used in the NRSA are
themselves direct indicators of human disturbance, and were used to screen reference sites, we did
not do t-tests to quantify the strength of relationship between W1_Hall2Lt\d general disturbance
class in Table 8.9. However, we do illustrate the relationship of W1_Hallto the human disturbance
gradient in Figure 8.3 to compare the relative magnitudes of W1_Hallamong least-, moderately-,
and most-disturbed streams in the various regions of the U.S.
We quantified the responsiveness of NRSA physical habitat condition metrics to levels of human
disturbance by the t-values (trt) of the difference between mean of the indicator Logio(0/E) values
in least-disturbed reference sites (prk3RRT_NRSA1314=K) minus the mean for the most-disturbed
sites (those screened as prkRRT_NRSA1314=T). Throughout the text, figures and tables, we
indicate the order of magnitude of p-values of these comparisons by the number of asterisks
following the t value. For example, trt = +2.34* indicates that the mean Logio(0/E) in reference
sites exceeds that in the most-disturbed sites by 2.34 log units, and = p~ 0.1. Multiple asterisks
118
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
denote the magnitude of p values (* = p~ 0.1; ** = p~ 0.01; *** = p~ 0.001; and **** = p~
0.0001).
Regional differences in bed substrate texture do not necessarily indicate anthropogenic
sedimentation. In other words, there are streams and rivers that are naturally fine-bedded.
Examination of the distribution of the Logio of geometric mean bed surface substrate diameter
('LSUB_dmm) shows that the wadeable streams clearly separate into fine-bedded and coarse-bedded
regions (Figure 8.4). Wadeable streams in CPL, UMW, and Central Plains (NPL, SPL, TPL) are
largely low gradient streams, and median bed sediments with LSUB_dmm <0 (<1 mm) which is
sand or finer (Figure 8.4b). A similar, but less distinct pattern is seen in boatable rivers, but NPL
and XER rivers are relatively more coarse-bedded than expected from the pattern in wadeable
streams. These patterns are driven largely by the slope and lithology of these sites. Patterns in the
distribution of LRBS_g08 (=LRBS_use) show less difference among regions, and a number of the
fine-bedded regions have similar bed stability as those found in coarser regions (Figure 8.4a and
Figure 8.4b). Once scaled as an O/E variable (LOE_LRBS_use) to adjust for natural controls on
bed material size and more clearly reflect anthropogenic influences, LKBS showed modest to
strong negative response to human disturbance for combined boatable and wadeable sites in most
regions and aggregations of regions, as illustrated by trt values ranging from +3.38*** to
+ 12.84****, showing substantial and statistically significant differences between means of least-
disturbed minus most-disturbed sites (Table 8.9). The strength of associations of instream
sediments with human disturbance (Table 8.9 and Figure 8.5) tended to be similar and relatively
strong for both boatable and wadeable rivers and streams (trt = +2.24** to +11.32****). We
observed moderate to strong declines LRBS with disturbance in all regions, the strongest
associations were in the UMW boatable sites (trt = 6.59 ****), the Western Rivers (trt = 3.96***), and
in EHIGH, CENPL, and WEST wadeable sites (trt = 4.12**** to 8.25****).
Riparian vegetation cover (LPt01 _XCMGW) adjacent to both wadeable and boatable rivers and
streams was markedly lower in the NPL than in any other region (Figure 8.6). By contrast, riparian
vegetation cover for both types of waters was consistently higher in the CPL, NAP, and SAP, with
the other regions having moderately high median values of riparian cover. Once scaled as an O/E
variable to adjust for natural geoclimatic controls (LOE_XCMGW_use), riparian vegetation cover
complexity showed modest to strong negative response to human disturbance for combined
boatable and wadeable sites in most regions and aggregations of regions, as illustrated by trt values
from +2.95*** to +14.17****); showing substantial and statistically significant differences between
means of least-disturbed minus most-disturbed sites (Table 8.9). Compared with the similar
response of sediment to disturbance in boatable and wadeable sites, the association between
riparian vegetation and disturbance was much stronger for wadeable sites (trt = +4.06**** to
+ 13.46****) than for boatable sites (tit = -0.13 to +3.44***) sites (Table 8.9 and Figure 8.7).
Among boatable rivers, riparian vegetation cover complexity was moderately correlated with the
disturbance levels only in the Coastal Plain (trt = 2.99***) and West (trt = 2.24**), and relatively
weakly associated elsewhere (tit = -0.13ns to 1.73*). Among wadeable streams, however, riparian
119
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
vegetation was strongly correlated with disturbance in all regions (trt = 4.06**** in the UMW to
7.35**** in CENPL). Note of course that expected riparian vegetation cover complexity is much
higher in the CPL and EHIGH, for example, than in the CENPL.
Instream habitat cover complexity (LPt01 _XFC_NAT) in boatable and wadeable rivers and
streams was markedly lower in the NPL than in any other region (Figure 8.8). In wadeable
streams, the Central Plains ecoregions (NPL, SPL, and TPL) had markedly lower instream cover
complexity than the other regions. Boatable and wadeable rivers and streams in the SAP, CPL, and
NAP, and wadeable rivers and streams in the WMT had generally higher instream habitat cover
complexity than the other regions (Figure 8.8). We scaled instream cover complexity as an O/E
variable ([LOE_XFC_NAT_use) to adjust for geoclimatic influences on instream cover, we
examined the associations between instream cover and anthropogenic influences (Table 8.9 and
Figure 8.9). Except for the weak response in the Upper Midwest (trt = +1.08*), the instream
habitat complexity indicator showed moderate response to human disturbance, with ta values
ranging from +2.30** to +6.62**** for combined boatable and wadeable sites (Table 8.9).
However, as was the case for the riparian vegetation indicators, associations were in most cases
much stronger for wadeable (trt = +1.73* to 8.16****), than for boatable sites (Figure 8.9), where
most regional associations of instream habitat complexity to human disturbance levels were non-
significant, with low or negative t values (-1.78* to +0.91*). Among wadeable sites, however, the
associations of instream habitat complexity with disturbance ranged from weak in the EHIGH and
UMW (trt = 1.73* and 2.02**) to very strong in the WEST (trt = 8.16 and p < 0.0001). Note that
expected instream habitat complexity is generally higher in the CPL and upland regions (EHIGH
and WEST) than for the CENPL and UMW.
8.5 Literature Cited
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid bioassessment protocols for use in
streams and wadeable rivers: periphyton, benthic macroinvertebrates, and fish. EPA/841-B-99-002, U.S.
Environmental Protection Agency, Washington, D.C.
Benson, B.J., and J.J. Magnuson. 1992. Spatial heterogeneity of littoral fish assemblages in lakes:
relation to species diversity and habitat structure. Canadian Journal of Fisheries and Aquatic
Sciences 49:1493-1500.
Bisson, P.A., J.L. Nielsen, R.A. Palmason, and L.E. Grove. 1982. A system of naming habitat types
in small streams, with examples of habitat utilization by salmonids during low stream flow.
Pages 62-73 In N. B. Armantrout, [editor]. Acquisition and utilisation of aquatic habitat inventory
information. Symposium Proceedings, October 28-30, 1981, Portland, Oregon. The Hague
Publishing, Billings, Montana.
Bryce, S.A., Lomnicky, G.A, and P.R. Kaufmann. 2010. Protecting Sediment-Sensitive Aquatic
Species in Mountain Streams through the Application of Biologically-Based Criteria
Streambed Sediment Criteria./. North American Benthological S oc. 29(2):657-672.
120
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Bryce, S.A., G.A. Lomnicky, P.R. Kaufmann, L.S. McAllister, and T.L. Ernst 2008. Development of
Biologically-Based Sediment Criteria in Mountain Streams of the Western United States. N.
Am. J. Fish. Manage. 28: 28:1714-1724.
Buffmgton, J.M., and D.R. Montgomery. 1999a. Effects of hydraulic roughness on surface textures
of gravel-bed rivers. Water Resources Research 35:3507-3521.
Buffmgton, J.M., and D.R. Montgomery. 1999b. Effects of sediment supply on surface textures of
gravel-bed rivers. Water Resources Research 35:3523-3530.
Cummins, K.W. 1974. Structure and function of stream ecosystems. BioScience 24:631-641.
Dietrich, W.E., J.W. Kirchner, H. Ikeda, and F. Iseya. 1989. Sediment supply and the development
of the coarse surface layer in gravel bed rivers. Nature 340:215-217.
Dingman, S.L. 1984. Fluvial Hydrology. W.H. Freeman, New York.
Dunne, T., and L.B. Leopold. 1978. Water in environmental planning. W. H. Freeman and Co., New
York
Faustini, J.M., and P.R. Kaufmann. 2007. Adequacy of visually classified particle count statistics
from regional stream habitat surveys. Journal of the American Water Resources Association
43:1293-1315.
Fitzpatrick, F.A., I.R. Waite, P.J. D'Arconte, M.R. Meador, M.A. Maupin, and M.E. Gurtz. 1998.
Revised methods for characterizing stream habitat in the National Water-Quality Assessment
Program. Water-Resources Investigations Report 98-4052, U.S. Geological Survey Reston, Virginia.
Frissell, C.A., W.J. Liss, C.E. Warren, and M.D. Hurley. 1986. A hierarchical framework for stream
habitat classification: viewing streams in a watershed context. Environmental Management
10:199-214.
Gordon, N.D., T.A. McMahon, and B.L. Finlayson. 1992. Stream hydrology, an introduction for ecologists.
John Wiley & Sons, New York.
Gorman, O.T., and J.R. Karr. 1978. Habitat structure and stream fish communities. Ecology 59:507-
515.
Gregory, S. V., F.J. Swanson, W.A. McKee, and K.W. Cummins. 1991. An Ecosystem Perspective
of Riparian Zones. BioScience 41:540-551.
Hankin, D.G., and G.H. Reeves. 1988. Estimating total fish abundance and total habitat area in
small streams based on visual estimation methods. Canadian Journal of Fisheries and Aquatic
Sciences 45:834-844.
Harmon, M.E., J.F. Franklin, F.J. Swanson, P. Sollins, S.V. Gregory, J.D. Lattin, N.H. Anderson,
S.P. Cline, N.G. Aumen, J.R. Sedell, G.W. Lienkamper, K. Cormack Jr., and K.W. Cummins.
1986. Ecology of Coarse Woody Debris in Temperate Ecosystems. Advances in Ecological
Research 15:133-302.
Harrelson, C.C., C.L. Rawlins, and J.P. Potyondy. 1994. Stream channel reference sites: an illustrated
guide to field technique. General Tech. Rp. RM-245, USDA Forest Service, Rocky Mountain
Forest and Range Experiment Station, Fort Collins, Colorado.
Hawkins, C.P., M.L. Murphy, and N.H. Anderson. 1983. Density of fish and salamanders in relation
to riparian canopy and physical habitat in streams of the northwestern United States.
Canadian Journal of Fisheries and Aquatic Sciences 40:1173-1186.
121
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Hawkins, C.P., J.L. Kershner, P.A. Bisson, M.D. Bryant, L.M. Decker, S.V. Gregory, D.A.
McCullough, C.K. Overton, G.H. Reeves, R.J. Steedman, and M.K. Young. 1993. A
Hierarchical Approach to Classifying Stream Habitat Features. Fisheries 18:3-12.
Herlihy, A.T., S.G. Paulsen., J. Van Sickle, J.L. Stoddard. C.P. Hawkins, and L.L. Yuan. 2008.
Striving for consistency in a national assessment: the challenges of applying a reference-
condition approach at a continental scale./. N. Am. Benthol Soc. 27(4):860877.
Hughes, R.M., and D.V. Peck. 2008. Acquiring data for large aquatic resource surveys: the art of
compromise among science, logistics, and reality. Journal of the North American Benthological
Society 27:837859.
Hynes, H.B.N. 1970. The ecology of running waters. Unversity of Toronto Press, Toronto, Ontario,
Canada.
Kappesser, G.B. 2002. A riffle stability index to evaluate sediment loading to streams. Journal of the
American Water Resources Association 38:1069-1081.
Kaufmann, P.R. 1987a. Channel morphology and hydraulic characteristics of torrent-impactedforest streams in the
Oregon Coast Range, U.S.A. Ph.D. Dissertation. Oregon State University, Corvallis, Oregon.
Kaufmann, P.R. 1987b. Slackwater habitat in torrent-impacted streams. Pages 407-408 In: R. L.
Beschta, T. Blinn, G.E. Grant, F.J. Swanson, and G.E. Ice, (Eds.). Erosion and
Sedimentation in the Pacific Rim. International Association of Hydrologic Science, Pub. No.
165, Proceedings of an International Symposium, August 3-7, 1986, Ore. State Univ., Corvallis, OR.
International Association of Hydrologic Science.
Kaufmann, P.R. 1993. Physical habitat. Pages 59-69 In R. M. Hughes. Stream Indicator and Design
Workshop. EPA 600/R-93/138, U.S. Environmental Protection Agency, Office of Research
and Development, Corvallis, Oregon.
Kaufmann, P.R., and J.M. Faustini, 2012. Simple measures of channel habitat complexity predict
transient hydraulic storage in streams. Hydrobiologia. 685:69-95.
Kaufmann, P.R., and R.M. Hughes. 2006. Geomorphic and Anthropogenic Influences on Fish and
Amphibians in Pacific Northwest Coastal Streams. In: R.M. Hughes, L. Wang and P.W.
Seelbach (Editors), Landscape Influences on Stream Habitats and Biological Assemblages.
American Fisheries Society Symposium 48:429-455, Bethesda, Maryland.
Kaufmann, P.R., and T.R. Whittier. 1997. Habitat Assessment. Pages 5-1 to 5-26 In J.R. Baker, D.V.
Peck, and D. W. Sutton. Environmental Monitoring and Assessment Program -Surface Waters: Field
Operations Manual for Takes. EPA/620/R-97/001, U.S. Environmental Protection Agency,
Washington, D.C.
Kaufmann, P.R., D.P. Larsen, and J.M. Faustini, 2009. Bed Stability and Sedimentation Associated
With Human Disturbances in Pacific Northwest Streams./. Am. Water Resources Assoc.
45(2):434-459.
Kaufmann, P.R., J.M. Faustini, D.P. Larsen, and M.A. Shirazi. 2008. A roughness-corrected index of
relative bed stability for regional stream surveys. Geomophology 199:150-170.
Kaufmann, P.R., P. Levine, E.G. Robison, C. Seeliger, and D.V. Peck. 1999. Quantifyingphysical
habitat in wadeable streams. EPA/620/R-99/003, U.S. Environmental Protection Agency,
Washington, D.C.
122
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Keim, R.F., A.E. Skaugset, and D.S. Bateman, 2002. Physical aquatic habitat II. Pools and cover
affected by large woody debris in three western Oregon streams. North American Journal of
Fisheries Management 22: 151164.
Kovalenko, K.E., S.M. Thomaz, and D.M. Warfe. 2012. Habitat complexity: approaches and future
directions editorial review. Hydrobiologia 685:117. DOI 10.1007/sl0750-011-0974-z
Larsen, D.P., P.R. Kaufmann, T.M. Kincaid, and N.S. Urquhart. 2004. Detecting persistent change
in the habitat of salmon-bearing streams in the Pacific Northwest. Canadian Journal of Fisheries
and Aquatic Sciences 61:283-291.
Lazorchak, J.M., D.J. Klemm, and D.V. Peck. 1998. 'Environmental Monitoring and Assessment Program-
Surface Waters: field operations and methods for measuring the ecological condition of wadeable streams.
EPA/620/R-94/004F, U.S. Environmental Protection Agency, Washington, D.C.
Leopold, L.B. 1994. A View of the River. Harvard University Press, Cambridge, Massachusetts.
Leopold, L.B., M.G. Wolman, and J.P. Miller. 1964. Fluvial processes in geomorphology. W.H. Freeman,
San Francisco.
Lisle, T.E. 1982. Effects of aggradation and degradation on riffle-pool morphology in natural gravel
channels, northwestern California. Water Resources Research 18:643-1651.
Lisle, T.E. 1987. Using "residual depths" to monitor pool depths independently of discharge.
Research Note PSW-394, USDA Forest Service, Pacific Southwest Forest and Range
Experimental Station, Berkeley, California.
Lisle, T.E., and S. Hilton. 1992. The volume of fine sediment in pools: an index of sediment supply
in gravel-bed streams. Water Resources Bulletin 28:371-383.
MacDonald, L.H., A.W. Smart, and R.C. Wismar. 1991. Monitoring Guidelines to Evaluate Effects of
F'orestry Activities on Streams in the Pacific Northwest and Alaska. EPA 910/9-91-001, U.S.
Environmental Protection Agency, Region X, Seattle, Washington.
Madej, M.A. 2001. Development of channel organization and roughness following sediment pulses
in single-thread, gravel bed rivers. Water Resources Research 37:2259-2272.
Montgomery, D.R., and J.M. Buffmgton. 1993. Channel classification, prediction of channel
response, and assessment of channel condition. Washington State Timber/Fish/Wildlife
Agreement, Report TFW-SH10-93-002, Department of Natural Resources, Olympia,
Washington.
Montgomery, D.R., and J.M. Buffmgton. 1997. Channel-reach morphology in mountain drainage
basins. Geological Society of America Bulletin 109.
Montgomery, D.R., and J.M. Buffmgton. 1998. Channel processes, classification, and response.
Pages 13-42 In R. Naiman and R. Bilby, [editors]. RJverEcology and Management. Springer-
Verlag, New York.
Moore, K. M.S., and S.V. Gregory. 1988. Summer habitat utilization and ecology of cutthroat trout
fry (Salmo clarki) in Cascade mountain streams. Canadian Journal of Fisheries and Aquatic S ciences
45:1921-1930.
Morisawa, M. 1968. Streams, their dynamics and morphology. McGraw-Hill Book Company, New York.
123
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Mossop, B. and M.J. Bradford, 2006. Using thalweg profiling to assess and monitor juvenile salmon
(Oncorhynchus spp.) habitat in small streams. Canadian Journal of Fisheries and Aquatic S ciences 63:
1515-1525.
Naiman, R.J., H. Decamps, J. Pastor, and C.A. Johnston. 1988. The potential importance of
boundaries to fluvial ecosystems. Journal of the North American Benthological Society 7:289-306.
O'Neill, M.P., and A.D. Abrahams. 1984. Objective identification of pools and riffles. Water Resources
Research 20:921-926.
Paulsen, S.G., A. Mayio, D.V. Peck, J.L. Stoddard, E. Tarquinio, S.M. Holdsworth, J. Van Sickle,
L.L. Yuan, C.P. Hawkins, A.T. Herlihy, P.R. Kaufmann, M.T. Barbour, D.P. Larsen, and
A.R. Olsen. 2008. Condition of stream ecosystems in the US: an overview of the first
national assessment. J. N. Am. BenthologicalSoc. 27(4):812821.
Pearsons, T.N., and G.M. Temple, 2007. Impacts of early stages of salmon supplementation and
reintroduction programs on three trout species. North American Journal of Fisheries Management
27: 1-20.
Pearsons, T.N., and G.M. Temple, 2010. Changes to Rainbow Trout abundance and salmonid
biomass in a Washington watershed as related to hatchery salmon supplementation.
Transactions of the American Fisheries Society 139: 502520.
Peck, D.V., D.K. Averill, A.T. Herlihy, B.H. Hill, R.M. Hughes, P.R. Kaufmann, D.J. Klemm, J.M.
Lazorchak, F.H. McCormick, S.A. Peterson, P.L. Ringold, M.R. Cappaert, T. Magee, and
P. A. Monaco. (in press) Environmental Monitoring and Assessment Program: Surface Waters Western
Pilot Study-field operations manualfor nonwadeable streams. EPA 620/ R-xx/xxx, U.S.
Environmental Protection Agency, Washington, D.C.
Peck, D.V., A.T. Herlihy, B.H. Hill, R.M. Hughes, P.R. Kaufmann, D.J. Klemm, J.M. Lazorchak,
F.H. McCormick, S.A. Peterson, P.L. Ringold, T. Magee, and M.R. Cappaert. 2006.
Environmental Monitoring and Assessment Program: Surface Waters Western Pilot Studyfield
operations manualfor wadeable streams. EPA/620/R-06/003, U.S. Environmental Protection
Agency, Washington, D.C.
Platts, W. S., W.F. Megahan, and G.W. Minshall. 1983. Methods for evaluating stream, riparian, and
biotic conditions. General Technical Rport INT-138, U.S. Department of Agriculture, Forest
Service, Intermountain Forest and Range Experiment Station, Ogden, Utah.
Rinne, J. 1988. Effects of livestock grazing exclosure on aquatic macroinvertebrates in a montane
stream, New Mexico. Great Basin Naturalist 48:146-153.
Robison, E.G., and R.L. Beschta. 1989. Estimating stream cross sectional area from wetted width
and thalweg depth. Physical Geography 10:190-198.
Robison, E.G. and P.R. Kaufmann. 1994. Evaluating two objective techniques to define pools in
small streams. Pages 659-668 In R. A. Marston and V. A. Hasfurther, [editors]. Effects of
Human Induced Changes on Hydrologic Systems. Summer Symposium proceedings, American
Water Resources Association. June 26-29, 1994, Jackson Hole, Wyo.
Stack, W.R., and R.L. Beschta. 1989. Factors influencing pool morphology in Oregon coastal
streams. Pages 401-411 In W. W. Woessner and D. F. Potts, [editors]. Headwaters Hydrology
Symposium. American Water Resources Association.
124
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Stoddard, J.L., D.V. Peck, A.R. Olsen, D.P. Larsen, J. Van Sickle, C.P. Hawkins, R.M. Hughes, T.R.
Whittier, G. Lomnicky, A.T. Herlihy, P.R. Kaufmann, S.A. Peterson, P.L. Ringold, S.G.
Paulsen, and R. Blair. 2005a. 'Environmental Monitoring and Assessment Program (EMAP): western
streams and rivers statistical summary. EPA 620/R-05/006, U.S. Environmental Protection
Agency, Washington, D.C.
Stoddard, J.L., D.V. Peck, S.G. Paulsen, J. Van Sickle, C.P. Hawkins, A.T. Herlihy, R.M. Hughes,
P.R. Kaufmann, D.P. Larsen, G. Lomnicky, A.R. Olsen, S.A. Peterson, P.L. Ringold, and
T.R. Whittier. 2005b. An ecological assessment of western streams and rivers. EPA 620/R-05/005,
U.S. Environmental Protection Agency, Washington, D.C.
Suttle, K.B., M.E. Power, J.M. Levine, and C. McNeely. 2004. How fine sediment in riverbeds
impairs growth and survival of juvenile salmonids. Ecological Applications 14:969974.
USEPA. 2004. Wadeable Streams Assessment: field operations manual. EPA/841/B-04/004, U.S.
Environmental Protection Agency, Washington, D.C.
USEPA. 2006. Wadeable Streams Assessment: a collaborative survey of the Nation's streams. EPA/641/B-
06/002, U.S. Environmental Protection Agency, Washington, D.C.
USEPA 2007. National Rivers and Streams Assessment: Field Operations Manual. EPA-841-B-07-009. U.S.
Environmental Protection Agency, Washington, D.C.
USEPA. 2013. National Rivers and Streams Assessment 2013-2014: Field Operations Manual Non-
Wadeable. EPA-841-B-12-009a. U.S. Environmental Protection Agency, Office of Water,
Washington, DC.
USEPA. 2013. National Rivers and Streams Assessment 2013-2014: Field Operations Manual Wadeable.
EPA-841-B-12-009b. U.S. Environmental Protection Agency, Office of Water, Washington,
DC.
USEPA. 2016. National Rivers and Streams Assessment 2008-2009: A Collaborative Survey. EPA
841-R-l 6-007. U.S. Environmental Protection Agency, Office of Water and Office of
Research and Development, Washington, DC.
Wilcock, P. R. 1997. The components of fractional transport rate. Water Resources Research 33:247-
258.
Wilcock, P. R. 1998. Two-fraction model of initial sediment motion in gravel-bed rivers. Science
280:410-412.
125
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.1 Metrics used to characterize the general attributes of stream/river physical habitat.
Habitat Volume:
LRP100 = log(RPlOO) = Log of Mean Residual Depth (cm)
Scaled Habitat Volume:
LDVRP100 = log(RPlOO) - log(PredictedRP100) = Deviation in Mean Residual Depth from expected value
Habitat Complexity:
CVDPTH = SDDEPTH /XDEPTH = Coefficient of Thalweg Depth Variation
CI WM100 = Number of Large Woody Debris pieces/lOOm of channel.
LV1 W_MSQ = log[Volume of Large Woody Debris per m2 of bankfull channel area (m3/m2)].
XFC_NA T = Areal Cover of Woody Debris, Brush, Undercut Banks, Overhanging Vegetation, plus Boulders and
Rock Ledges.
XFC NORK = Areal Cover of Woody Debris, Brush, Undercut Banks, Overhanging Veg.
XFCAQM = Areal Cover of Aquatic Macrophytes
XFC ALG = Areal Cover of Filamentous Algae detectable by the unaided eye.
Streambed Particle Size:
LSUB dmm = log[Streambed surface particle - mm] = log of geometric mean diameter of bed surface
sediments in millimeters.
PCT_FN = % Streambed Silt & Finer
PCT_SAFN = % Streambed Sand & Finer
XEMBED = % Substrate Embedded by Sand and Fines
Scaled Streambed Particle Size:
DPCT FN = Deviation of PCT FN from expected value ("excess Fines")
DPCT SF = Deviation of PCT_SAFN from expected value ("excess Sand+Fines")
DEVLSUB = Deviation of LSUB DMM from expected value (Streambed Fining Index)
Relative Bed Stability:
LRBS= logio of diameter ratio: Geometric mean bed particle diameter / Critical (mobile) diameter at bankfull flow
stage. (LRBS_bw5: see Kaufmani! el ai 1999; LRBS_g08: see Kaufmani! el at. 2008, 2009).
Floodplain Interaction:
LSINU = Log(SINU) = Log(Channel Sinuosity).
L1NC1SH = log(XINC_H - XBKF H + 0.1) = Log of Incision from terrace to bankfull ht (m).
LBFWDRAT = log{BKF_W/BKF H+ (XDEPTH/100)}= log (Bankfull Width/Depth Ratio)
LBFXWRAT = log(BKF_W/XWIDTH) = log (Bankfull Width / Wetted Width) (an index of streamside flood
inundation potential)
Hydrologic Regime:
LQSLTR RAT = log{(Qsp+0.000000l)/LTROFF_M}= log{low flow /annual mean runoff} (~ an inverse index of
"droughtiness",
where: Qsp = Flow_mps/WSAREAKM= (flow_cfs/35.315)/WSAREAKM
LBFXDRA T =log{(XBKF_H+ (XDEPTH/100) / (XDEPTH/100)} = log(ratio of bankfull depth / wetted depth), a
morphometric index of "flashiness".
Riparian Vegetation:
XCDENMID: % Canopy Density measured midstream.
XCMG = Riparian Canopy+Mid-+Ground Layer Vegetation (areal cover proportion)
XCMG W = Riparian Canopy+Mid+Ground Layer Woody Veg. (areal cover proportion)
Riparian Habitat Alteration:
QR 1=(QRVEG 1 *QRVEG2*QRDISTl)a3333; where:
if XCMGW <=2.00 then QRVegl=A+(Q.9(XCMGW/2.QQ));
ifXCMGW>2.00 then QRVegl=1;
QRVeg2=A+(0.9(XCDENBK/100)); and QRD1ST1 = 1/(1+W1_HALL )
Riparian Human Disturbances:
W1HAG = Riparian & near-Stream Agriculture - all types (proximity-weighted tally)
W1H ROAD = Riparian & near-Stream Roads (proximity-weighted tally)
W1H CROP = Riparian & near-Stream Row Crop Agriculture (proximity-weighted tally)
W1H WALL = Riparian & near-Stream Walls, Dikes, Revetment (proximity-weighted tally)
W1HALL = Proximity-weighted Index of Human Disturbances of All Types
QRD1ST1 = \!{\+Wl_HALL ) = Proximity-weighted Inverse Index of Human Disturbances of All Types
126
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.2 Sampling revisit precision (repeatability) of the four physical habitat condition indicators.
Repeat visits within the summer sampling season were used to calculate RMSrep, which is essentially the
standard deviation of repeat sampling pairs to the same stream or river reach. Dividing the square of the
RMSrep into the variance among sites gives the S:N variance ratio. (See Kaufmann etal. 1999 for ANOVA
methods to calculate RMSrep and S:N, where RMSrep is equal to their RMSE.)
Metric
GrouD
Sites (n)
mean
ReDeat
Dairs (n)
RMSren
S:N
LRBS_g08
All Sites
4058
-0.938
375
0.519
4.17
All
(0809/ 1314)
(2032/2025)
(-0.942/-0.933)
(191/184)
(0.482/0.556)
(5.13/3.42)
Boatable
1484
-0.661
178
0.479
6.70
Wadeable
2573
-1.104
197
0.553
2.58
EHIGH
1075
-0.541
134
0.539
3.65
PLNLOW
2060
-1.242
164
0.514
3.89
WMTNS
921
-0.740
77
0.493
4.07
L_xcmgw
All Sites
4193
-0.252
388
0.148
8.45
All
(0809/ 1314)
(2112/2080)
(-0.286/-0.218)
(197/191)
(0.146/0.150)
(9.38/7.46)
Boatable
1599
-0.154
187
0.144
4.70
Wadeable
2593
-0.315
201
0.151
10.08
EHIGH
1100
-0.051
138
0.083
8.05
PLNLOW
2158
-0.341
173
0.188
6.72
WMTNS
933
-0.293
77
0.135
7.79
L_xfc_nat
All Sites
4193
-0.603
388
0.214
2.27
All
(0809/ 1314)
(2112/2080)
(-0.590/-0.617)
(197/191)
(0.240/0.184)
(1.87/2.99)
Boatable
1599
-0.626
187
0.220
1.66
Wadeable
2593
-0.589
201
0.209
2.76
EHIGH
1100
-0.494
138
0.200
1.57
PLNLOW
2158
-0.670
173
0.227
2.24
WMTNS
933
-0.584
77
0.211
2.28
L_Wl_Hall
All Sites
4193
-0.129
388
0.178
5.46
All
(0809/ 1314)
(2112/2080)
(-0.152/-0.106)
(197/191)
(0.186/0.170)
(5.18/5.76)
Boatable
1599
-0.091
187
0.137
9.03
Wadeable
2593
-0.154
201
0.210
3.89
EHIGH
1100
-0.078
138
0.181
5.15
PLNLOW
2158
-0.151
173
0.168
5.85
WMTNS
933
-0.142
77
0.196
5.10
127
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.3 Estimated number of years to detect trends in habitat attributes. Number of years required for a
50-site monitoring network to detect 1% and 2% per year trends in habitat attributes with 80% likelihood
(beta, or power) and alpha = 0.05, if specified trends occur, and sites are visited each year. Data were
taken from Larsen et al. (2004),a or calculated using the same data and analytical procedures used in that
publication.13
Variable
Description
1% trend
2% trend
SDDEPTHh
(Std. Deviation of Thalweg Depth)
13 years
8 years
LRP100*
(log[Mean Residual Depth])
20
12
PCT_SAFNa
(% Sand + Silt)
21
13
XEMBEiy
(% Embeddedness)
20
12
LRBS_Bmi
(log[Rel. Bed Stability])
12
8
LV1W_MSQ
(logfLarge Wood Volume/m2])
27
17
XCMGWh
(3-Layer Riparian Woody Veg Areal Cover)
12
8
XCDENMID"
(Canopy Density measured midstream)
13
8
128
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.4 Anthropogenic disturbance screening criteria
Criteria used.to characterize least-disturbed reference (R), moderately-disturbed (S), and most-disturbed (T) sample reaches for developing
physical habitat condition criteria. In addition to the tabulated criteria, potential reference sites were rejected if DAM_DII > 1, or
URB_1 KMCIRCLE > 5%, or AG_1KMCIRCLE > 15%.
Values > than those before the slash (/) are EXCLUSION criteria for least-disturbed reference sites.
Values > those after slash are INCLUSION criteria for most-disturbed sites.
W, B, and G refer to Wadeable, Boatable, and Great River sites.
Region
PTL
NTL
a
SQ4
Turb
W1 HALL
W1 HAG
Wadeable
W1H CROP
Wadeable
Wll I WALL
Wadeable
NAP
20/100
750/3500
250/10000
250/1000
5/10
2.0/4.0
0.1/0.4
0.05/0.10
0.2/0.4
SAP
20/100
750/3500
200/1000
400/1000
5/20
2.0/4.0
0.1/0.4
0.05/0.10
0.2/0.4
UMW
50/150
1000/5000
300/2000
400/2000
5/30
2.0/4.0
0.15/1.4
0.1/0.4
0.2/0.4
CPL
75/250
2500/8000
999999/
999999
600/4000
10/50
2.0/4.0
0.15/1.4
0.05/ 0.4
0.2/0.4
TPL
100/500
3000/15000
2000/5000
999999/
999999
50/100
2.0/4.0
0.67/1.4
0.25/0.48
0.4/0.6
NPL&
SPL
150/500
4500/10000
1000/5000
999999/
999999
50/100
2.0/3.0
1.0/1.4
0.15/ 0.25
0.2/0.4
WMT:
Southwest
50/100
750/1500
300/1000
99999/
99999
5/10
W:0.5/3.0
B,G:1.5/3.0
0.25/1.4
0.10/0.25
0.2/0.4
S.Rockies
25/100
750/1500
200/1000
200/1000
5/10
W:l.0/3.0
B,G:1.5/3.0
0.3/1.4
0.1/0.25
0.2/0.4
N.Rockies
& Pacific
25/100
750/1500
200/1000
200/1000
5/10
W:0.5/3.0
B,G:1.5/3.0
0.3/1.4
0.10/0.25
0.2/0.4
XER
50/150
1500/5000
1000/5000
999999/
999999
25/75
1.5/3.0
0.6/1.4
0.15/0.25
0.2/0.4
129
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.5 NRSA boatable and wadeable least-disturbed reference sites from combined 2008-09 & 2013-14
surveys, selected using consistent criteria listed in Table 8.4. Numbers of reference sites identified from
the 2008-09 and 2013-14 surveys are parenthesized and separated by a slash (/).
ECQ9 ECOp5 Total Boatable Wadeable
NAP APPAL 88 (45/43) 47 (24/23) 41 (21/20)
SAP APPAL 54_(40Z14)_ 22_(15/7) _32_ (25/7)
CPL _CPL_ 103155/48) 52125/27j_ _ _ _5_1_ (3_0J 21)
UMW_ _UMW _79_(40739) 36_(18 /18) _ _ _ _4_3_ (2_2J 21)
TPL CENPL 83 (44/39) 22 (12/10) 61 (32/29)
NPL CENPL 85 (29/56) 33 (11/22) 52 (18/34)
SPL_ _CENPL _44 (2_3/2_1) 2_l2/0) _4_2_Qljll)
WMT WEST 112 (47/65) 43 (16/27) 69 (31/38)
XER_ WEST 60_(26/34)_ 24_C6/18) 36^20/16^ _ _ _
Totals for lower 48 states 708 (349/359) 281(129/152) 427 (220/207)
130
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.6 Summary of regression models used in estimating site-specific expected values of LoglO
Relative Bed Stability (LRBS_g08) under least-disturbed reference conditions. See Appendix 8.A for model
details.
REGION/Realm
NAP/Boatable
ExpRef LRBS_g08 = (mean)Ref-4? (R2=0%, RMSE=1.539)
NAP/Wadeable
ExpRef LRBS_g08=f(LAws, Wl_HAG)RฃF-4i, where W1_HAG=0 (R2=22%, RMSE=0.525)
SAP/Boatable
ExpRef LRBS_g08 = (mean)RฃF-22 (R2=0%, RMSE=0.704)
SAP/Wadeable
ExpRef LRBS_g08=f(LAws, Wl_Hall)RฃF-32 where Wl_Hall=0 (R2=28%, RMSE=0.691)
CPL/Boatable
ExpRef LRBS_g08 = (mean)REF.52 (R2=0%, RMSE=1.331)
CPL/Wadeable
ExpRef LRBS_g08=f(LSIope, LWidth, Wl_Hall)REF-5i where Wl_Hall=0 (R2=35%, RMSE=0 .736)
UMW/Boatable
ExpRef LRBS_g08 = (Lot, Wl_Hall)REF-36 where Wl_Hall=0 (R2=18%, RMSE=1.259)
UMW/Wadeable
ExpRef LRBS_g08 = (LSIope, Wl_Hall)REF^3 where Wl_Hall=0 (R2=41%, RMSE=0.925)
NPL/Boatable
Exp LRBS_g08 =f(LAws, LSIope, [AGws-x-KFct])ALL-si, where AGws-x-KFct =0 (R2=56%, RMSE=0.610)
ExpRef (LRBS_g08/Exp LRBS_g08) =f(PCT_AG_WS)REF-2s where PCT_AG_WS =0 (R2=23%, RMSE=0.512)
NPL/Wadeable
Exp LRBS_g08 =f(Elev, LSIope, LWidth, Wl_Hall, Wl_Crop)ALL.314 (R2=39%, RMSE=0.837)
where Wl_Hall, Wl_Crop [AGws-x-KFct]) = 0
ExpRef (LRBS_g08/Exp LRBS_g08) =f(Wl_Hall)RฃF.5i (R2=3%, RMSE=0.839)
where Wl_Hall=0
SPL+TPL/Boatable
Exp LRBS_g08 =f(LAws, AG_lKMCircle)RฃF-47 (sputpunpl) where AG_lKMCirle =0 (R2=18%, RMSE=1.139)
SPL/Wadeable
Exp LRBS_g08 =f(Lat, LAws, LSIope, W1_HAG, AG_lKMCircle)ALL-297 (R2=35%, RMSE=0.952)
where W1_HAG, AG_lKMCircle = 0
ExpRef (LRBS_g08/Exp LRBS_g08) =f(WlH_NOAG, Dam_dii, RdDen_ws, PCT_AG_ws)REF^2 (R2=26%, RMSE=0.990)
where W1H_N0AG, Dam_dii, RdDen_ws, PCT_AG_ws = 0
TPL/Wadeable
Exp LRBS_g08 =f(Lat, Lon, LSIope)ALL.342 (R2=20%, RMSE=0.976)
ExpRef (LRBS_g08/Exp LRBS_g08) =
f(WlH_NOAG, WlH_Crop, AG_lKMCircle, PCT_AG_ WS, AgWS-x-KFct)RฃF-58 (R2=26%, RMSE=0.990)
where WlH_NOAG,WlH_Crop, AG_lKMCircle, PCT_AG_WS, AgWS-x-KFct = 0
WMT/Boatable
ExpRef LRBS_g08 = (mean)REF^3 (R2=0%, RMSE=0.365)
WMT/Wadeable
ExpRef LRBS_g08 =f(LSIope, LWidth)REF^9, (R2=27%, RMSE=0.430)
XER/Boatable
ExpRef LRBS_g08 = (mean)REF.24 (R2=0%, RMSE=0.985)
XER/Wadeable
ExpRef LRBS_g08 =f(LWidth)RฃF-36, (R2=23%, RMSE=0.794)
131
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.7 Summary of regression models used in estimating site-specific expected values of Riparian
Vegetation Cover and Structure (Logl0[0.01+XCMGW]) under least-disturbed reference conditions. See
Appendix 8.A for model details.
REGION/Realm
NAP/Boatable
ExpRef L_XCMGW =f(Lat,, AG_lKMCircle, PCT_AG_WS, AgWS-x-KFct)REF.47
where AG_lKMCircle, PCT_AG_WS, AgWS-x-KFct = 0
NAP/Wadeable
ExpRef L_XCMGW =f(LAws, LWidth, Wl_Hall)Ref-ai , where Wl_Hall=0
where W1 HAG=0
SAP/Boatable
ExpRef L_XCMGW =f(Wl_HAG)REF-22,
SAP/Wadeable
ExpRef L_XCMGW=f(LAws, ELEV, Wl_Ha\\)REf.32, where Wl_Hall=0
CPL/Boatable
ExpRef L_XCMGW=f(Lon, LAws, Wl_HAG)REF-52, where W1_HAG=0
CPL/Wadeable
ExpRef L_XCMGW =f(Lon)REF-5i
UMW/Boatable
ExpRef L_XCMGW = f(Lat, LAws, LSIope, LWidth)REfss (spl+tpl+umw)
UMW/Wadeable
ExpRef L_XCMGW=f(LSIope, LWidth, Wl_Hall)REF.43, where Wl_Hall=0
NPL+SPL /Boatable
Exp L_XCMGW =f(Lat, Lor1, W1JHAG, RDDEN_ws, PCT_AG_ws)all -249 (NPL+SPL+TPL)
where W1_HAG, RDDEN_ws, PCT_AG_ws = 0
ExpRef (L_XCMGW/Exp L_XCMGW) =f(PCT_AG_WS)REF-2s, where PCT_AG_WS =0
TPL/Boatable
ExpRef L_XCMGW = (mean)RCF-22
(R2=40%, RMSE=0.156)
(R2=24%, RMSE=0.121)
(R2=17%, RMSE=0.141)
(R2=32%, RMSE=0.141)
(R2=26%, RMSE=0.119)
(R2=l%, RMSE=0 .152)
(R2=34%, RMSE=0.373)
(R2=33%, RMSE=0.130)
(R2=25%, RMSE=0.362)
(R2=31%, RMSE=0.324)
(R2= 0%, RMSE=0.159)
(R2=31%, RMSE=0.487)
NPL/Wadeable
Exp L_XCMGW =f(Lat, Lor1, LSIope, LWidth, W1JHAG, PCT_AG_ws)all -922 (NPL+SPL+TPL)
where W1_HAG, PCT_AG_ws = 0
ExpRef (L_XCMGW/Exp L_XCMGW) =f(Damm_dii, PCT_AG_ws, AgWs-x-KFct)REF-is2 (npl+spl+tpl) (R2=14%, RMSE=0.386)
where Damm_dii, PCT_AG_ws, AgWs-x-KFct = 0
SPL+TPL/Wadeable
ExpRef LJXCMGW =f(Lon, ELEV, AG_lKMCircle, PCT_AG_ws, AGws-x-KFct)REF -143 (SPL+TPL+UMW) R2=40%, RMSE=0.267)
where AG_lKMCircle, PCT_AG_ws, AGws-x-KFct = 0
WMT/Boatable
ExpRef L_XCMGW = (mean)REF-43
WMT/Wadeable
ExpRef LJXCMGW =f(LAws, ELEV, LSIope,)REF-6s,
XER/Boatable
ExpRef L_XCMGW =f(Wl_HNOAG, Wl_HAG)REF-24,
XER/Wadeable
ExpRef L_XCMGW =f(LAws, LSIope, LWidth)REF-36,
where Wl_HNOAG, W1_HAG = 0
(R2= 0%, RMSE=0.262)
(R2=20%, RMSE=0.153)
(R2=29%, RMSE=0.153)
(R2=23%, RMSE=0.253)
132
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.8 Summary of regression models used in estimating site-specific expected values of Instream
Habitat Cover Complexity (Logl0[0.01+XFC_NAT]) under least-disturbed reference conditions. See
Appendix 8.A for model details.
REGION/Realm
NAP/Boatable
ExpRef L_XFC_NAT=f(Lon, LAws, LWidth, WlH_Crop)REF-47, where WlHJCrop = 0 (R2=34%, RMSE=0.319)
NAP/Wadeable
ExpRef L_XFC_NAT=f(LWidth)RฃF-4i (R2= 7%, RMSE=0.285)
SAP/Boatable
ExpRef L_XFC_NAT=f(Lat, Wl_Ha\\)REF.12, where Wl_Hall = 0 (R2=53%, RMSE=0.175)
SAP/Wadeable
ExpRef L_XFC_NAT=f(Lot, ELEV, W1_HAG)REf-32, where W1_HAG =0 (R2=42%, RMSE=0.310)
CPL/Boatable
ExpRef L_XFC_NAT = (mean)REF.52 (R2= 0%, RMSE=0.235)
CPL/Wadeable
ExpRef L_XFC_NAT = (mean)REF-5i (R2= 0%, RMSE=0.298)
UMW/Boatable
ExpRef L_XFC_NAT=f(Lon, W1_HAG)REF.36, where W1_HAG =0 (R2=23%, RMSE=0.316)
UMW/Wadeable
ExpRef L_XFC_NAT=f(LAws, LWidth)REF^3, (R2= 7%, RMSE=0.290)
NPL+SPL+TPL/Boatable
Exp L_XFC_NAT=f(Lat, Lor), LAws, ELEV, AG_lKMCircle)REF^7(NPusPUTPLi (R2=34%, RMSE=0.323)
where AG_lKMCircle = 0
NPL+SPL+TPL/Wadeable
Exp L_XFC_NAT=f(Lon, LAws, ELEV, AG_lKMCircle, URB_lKMCircle)REF -152 (NPL+SPL+TPL) (R2=17%, RMSE=0.335)
where AG_lKMCircle, URB_lKMCircle = 0
WMT/Boatable
ExpRef L_XFC_NAT=f(LWidth, WlH_Crop, RDDEN_ws)REM3, (R2= 24%, RMSE=0.230)
where WlH_Crop, RDDEN_ws = 0
WMT/Wadeable
ExpRef L_XFC_NAT=f(Lat, Lon, LAws, W1_HAG, RDDEN_ws)ref.6s, (R2=35%, RMSE=0.217)
where W1_HAG, RDDEN_ws = 0
XER/Boatable
ExpRef L_XFC_NAT=f(ELEV, LWidth)REF.23, (R2=13%, RMSE=0.310)
XER/Wadeable
ExpRef L_XFC_NAT=f(Lon, LSIope, WlH_Crop)REF.36, where WlHJCrop = 0 (R2=27%, RMSE=0.242)
133
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 8.9 Responsiveness to levels of human disturbance
Responsiveness of NRSA physical habitat condition metrics to levels of human disturbance, as quantified by t-values of the difference
between means of least-disturbed reference sites (prk3RRT_NRSA1314=R) minus most-disturbed sites (those screened as
prkRRT_NRSA1314=T). Values shown in red have a sign contrary to expectations. Order of magnitude of p-values shown by number
of asterisks (e.g., * = p~ 0.1; **** = p~ 0.0001)
Metric
Region
t-value R-T
(Boatable)
t-value R-T
(Wadeable)
t-value R-T
(All sites)
LOE_RBS_g08
USA-48
+11 32****
_l_7
+12 84****
CPL
+2.47**
+2.68**
+3.38***
EHIGH (NAP+SAP)
+3.22***
_l_4 ^2****
_l_4 yc)****
UMW
+6 59****
+2.24**
+5 32****
CENPL (TPL+NPL+SPL)
+2.64**
+6 39****
+6 93****
West (WMT+XER)
+3 96****
+8 25****
+8 98****
LOEXCMGW
USA-48
+3.44***
+13.46****
+14 17****
CPL
+2 99***
+5 69****
+6 32****
EHIGH (NAP+SAP)
+1.73*
+5 61****
+5 25****
UMW
-0.13
+4.06****
+2 95***
CENPL (TPL+NPL+SPL)
+1.43
+7 35****
+7 95****
West (WMT+XER)
+2.24**
+7 16****
+7 35****
LOE_XF C_N at
USA-48
-0.64
_l_7 g^****
+6.62****
CPL
+0.59
+3.64****
+3 52***
EHIGH (NAP+SAP)
+0.91*
+1.73*
+2.30**
UMW
-0.85
+2.02**
+1.08*
CENPL (TPL+NPL+SPL)
+0.56
+3 68****
+2.82**
West (WMT+XER)
-1.78*
+8 16****
+5 37****
134
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.1 Sample sites for NRSA 2008-09 and NRSA 2013-14.
A MRS A 2008-09 sample sites
B. NRSA 2013-14 sample sites
135
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.2. Riparian Disturbance (Wl_Hall) in combined NRSA 2008-09 and 2013-14 sample sites in 9
aggregate ecoregions of the conterminous U.S. Boxplots show 5th, 25th, median, 75th, and 95th percentiles of
the unweighted sample distributions (not population estimates). A. Boatable sites; B. Wadeable sites.
A. Boatable
W1JHALL
8
7
6
5
4
3
2
1
0
B. Wadeable
W1_HALL
8
7
6
5
D
4-....
2-
1-
CPL NAP NPL SAP SPL TPL UMW WMT XER
ECOWSA9 2015
a
0
1
CPL NAP NPL
ฆo-
o
6 ?
SAP SPL TPL UMW
ECOWSA9 2015
WMT XER
136
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.3 Riparian Disturbance (Wl_Hall) in combined NRSA 2008-09 and 2013-14 sample sites in 9
aggregate ecoregions of the conterminous U.S., contrasting distributions in least-, moderately-, and most-
disturbed sites within each aggregated ecoregion. Boxplots show 5th, 25th, median, 75th, and 95" percentiles
of the unweighted \sample distributions (not population estimates). A. Boatable sites; B. Wadeable sites.
Boatable EHiGH
W1 HALL
CPL
W1 HALL
UMW
W1 HALL
CENPL
W1 HALL
WEST
W1 HALL
1
prk3RRT... RSA1314
prk3RRT... RSA1314
prkSRRT... RSA1314
prk3RRT... RSA1314
prkSRRT... RSA'1314
Wadeable EHIGH
W1 HALL
"8
o
i f
5
c
\
1
1 T
8
ฃ l
R S T
prkSRRT... RSA1314
CPL
W1 HALL
7
6
5-
4
3-
2 o
o
J
R
o
o
8
~T~
S
8
I
~v
T
UMW
W1 HALL
CENPL
W1JHALL
8-
WEST
W1_HALL
8-'
~ E
prk3RRT...RSA1314
prk3RRT... RSA1314
prkSRRT... RSA1314
prk3RRT...RฃA1314
137
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.4 Log Relative Bed Stability (LRBS_use) and LoglO geometric mean bed surface substrate
diameter (LSUB dmm) in combined NRSA 2008-09 and 2013-14 sample sites in 9 aggregate ecoregions of
the conterminous U.S. Boxplots show 5th, 25th, median, 75th, and 95th percentiles of the unweighted sample
distributions (not population estimates). A. Boatable sites; B. Wadeable sites.
Boatable
LRBS use LSUB DMM
CPL NAP NPL SAP SPL TPL UMW WMT XER
ECOWSA9_2015
~ LRBS_use HLSUB_DMM
Wadeable
LSUB DMM
LRBS use
ฅ
ฃ ?
$
I
SAP SPL TPL
ECOWSA9 2015
UMW WMT
~ LRBS_use HLSUB_DMM
138
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.5 Observed/Expected Relative Bed Stability (LOE_LRBS_use) in combined NRSA 2008-09 and
2013-14 sample sites in 9 aggregate ecoregions of the conterminous U.S., contrasting distributions in least-
, moderately-, and most-disturbed sites within each aggregated ecoregion.Boxplots show 5th, 25th, median,
75th, and 95th percentiles of the unweighted sample distributions (not population estimates). A. Boatable
sites; B. Wadeable sites.
Boatable EHIGH CPL
trt=+3.22*** trt=+2.47**
LOE_RBS_use
4-
LOE_RBS_use
5
UMW
trt=+6.59****
LOE_RBS_use
CENPL WEST
trt=+2.64** trt=+3.96****
LOE_RBS_use LOE_RBS_use
3
2-
1
0
-1-
-2-
-3
-4-
I
R
I
S
~r
T
prk3RRT...RSA1314
prk3RRT...RSA1314
prk3RRT...RSA1314
prkSRRT...RSA1314
prk3RRT...RSA1314
Wadeable EHIGH CPL
trt=+4.12**** trt=+2.68**
UMW
trt=+2.24**
CENPL
trt=+6.39****
WEST
trt=+8.25****
139
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
LOE_RBS_use
3-
LOE_RBS_use
3^
LOE_RBS_use
3-
-2-
LOE_RBS_use
4-j:
LOE_RBS_use
6-
- 0
?
prk3 RRT... RSA1314
prk3RRT...RSA1314
prk3RRT... RSA1314
prk3RRT... RSA1314
prk3RRT...RSA1314
Figure 8.6 Riparian Vegetation Cover Complexity (LPt01_XCMGW) in combined NRSA 2008-09 and 2013-
14 sample sites in 9 aggregate ecoregions of the conterminous U.S. Boxplots show 5th, 25th, median, 75th,
and 95th percentiles of the unweighted sample distributions (not population estimates). A. Boatable sites;
B. Wadeable sites.
Boatable
Lpt01_XCMGW
0.5
0.0
-0.5
-1.0
-1.5
CPL NAP NPL
SAP SPL TPL
ECOWSA9 2015
t J
o 0
O ft
r ฅ -
1 *
L ! ,
O
ja 1 !
ฐ ^ ฐ
o ฐ
o o a
O O o
ฆW ?
1 ~T
o o
i ^ t 1'
UMW WMT
XER
Wadeable
140
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Lpt01_XCMGW
0.5
0.0
-0.5
-1.0
-1.5
-2.0
CPL NAP NPL SAP SPL TPL UMW WMT XER
ECOWSA9_2Q15
141
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.7 Observed/Expected Riparian Vegetation Cover Complexity (LOE_XCMGW_use) in combined
NRSA 2008-09 and 2013-14 sample sites in 9 aggregate ecoregions of the conterminous U.S., contrasting
distributions in least-, moderately-, and most-disturbed sites within each aggregated ecoregion. Boxplots
show 5th, 25th, median, 75th, and 95th percentiles of the unweighted sample distributions (not population
estimates). A. Boatable sites; B. Wadeable sites.
Boatable EHIGH
trt=+1.73*
LOE_XCMGW_use
0.5
0.0
-0.5
-1.0
-1.5
-2.0-
-2.5
i
R
I
S
I
T
prk3RRT... RSA1314
Wadeable EHIGH
trt=+5.61****
CPL
trt=+2.99***
LOE_XCMGW_use
0.5-
prk3RRT...RSA1314
CPL
trt=+5.69****
UMW
trt= -0.13n.s.
LOE_XCMGW_use
1.0
0.5-
0.0-
-0.5-
-1.0
1
-1.5-
-2.0
I
i i r
R S T
prk3RRT...RSA1314
UMW
trt=+4.06****
CENPL
trt=+1.43n.s.
LOE_XCMGW_use
0.5-
-1.5-
-2.0
0
0
h r
R S T
prk3RRT...RSA1314
CENPL
trt=+7.35****
WEST
trt=+2.24**
LOE_XCMGW_use
0.5-
0.0
-0.5-
-1.0-
-1.5-
-2.0
n i i
R S T
prk3RRT...RSA1314
WEST
trt=+7.16****
142
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
LOE_XCMGW_use
0.5-
0.0 -
J
-1.5 -
-2.0
-2.5-
-0.5- 0
-1.0
1
8
ฉ
o
o
"i 1 r
R S T
prk3RRT...RSA1314
LO E_XCMGW_u se
0.5
0.0-
-0.5-
-1.0
-1.5
-2.0
-2.5
LOE_XCMGW_use
0.5
0.0
-0.5
-1.0
-1.5-1
-2.0
1 1 1
-H 1 1 o
j ฐ
\ i
o
o
8
o
.8.
LO E_XCMGW_u se
1.0-
LO E_XCM GW_u s e
prk3RRT... RSA1314
R S T
prk3RRT... RSA1314
i
prk3RRT...RSA1314
prk3RRT...RSA1314
143
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.8 Instream Habitat Complexity (LPt01_XFC_NAT) in combined NRSA 2008-09 and 2013-14
sample sites in 9 aggregate ecoregions of the conterminous U.S. Boxplots show 5th, 25th, median, 75th, and
95th percentiles of the unweighted sample distributions (not population estimates). A. Boatable sites; B.
Wadeable sites.
Boatable
Lpt01_XFC_NAT
0.5
0.0
-0.5
-1.0
-1.5
-2.0
Wadeable
Lpt01_XFC_NAT
0.5
0.0
-0.5
-1.0
-1.5
-2.0
CPL NAP NPL SAP SPL TPL UMW WMT XER
ECOWSA9 2015
5 ฎ
6 o
CPL NAP NPL SAP SPL TPL UMW WMT XER
ECOWSA9 2015
144
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Figure 8.9 Observed/Expected Instream Habitat Complexity (LOE_XFC_NAT_use) in combined NRSA 2008-09
and 2013-14 sample sites in 9 aggregate ecoregions of the conterminous U.S., contrasting distributions in least-,
moderately-, and most-disturbed sites within each aggregated ecoregion. Boxplots show 5th, 25th, median, 75th,
and 95th percentiles of the unweighted sample distributions (not population estimates). A. Boatable sites; B.
Wadeable sites.
Boatable EHIGH
trt=+0.91*
LOE_XFC_NAT_use
1.5
-2.0 ฆ
CPL UMW
trt=+0.59ns- trt= -0.85ns-
LOE_XFC_NAT_u se LO E_XFC_NAT_use
i.o i.o-r
0.5-
o.o -
-0.5-
-1.0
-1.5 4
0.5-
o.o-
-0.5-
-1.0
CENPL
trt=+0.56ns-
LO E_XFC_N AT_use
1.0
0.5-
0.0 -rh
-0.5-
-1.0
-1.5-
-2.0
WEST
trt= -0.78ns-
LO E_XFC_N AT_u se
1.0
0.5-
0.0 -rh
-0.5-
-1.0
-1.5-
-2.0
prkSRRT...RSA1314
prk3RRT...RSA1314
prk3RRT...RSA1314
prk3RRT...RSA1314
prkSRRT...RSA1314
Wadeable EHIGH
trt=+1.73*
CPL
trt=+3.64****
UMW
trt=+2.02**
CENPL
trt=+3.68****
WEST
trt=+8.76****
145
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
LOE_XFC_NATjJse
1.0
LOE_XFC_NAT_use
1.0
0.5
1
o.o-
-1.5-
-2.0
-0.5-
-1.0-
o o
. o p
o
LO E_XFC_NAT_use LO E_XFC_NAT_use
1.0 H
0.5-
~i 1 r
R S T
prk3RRT.. RSA1314 prkSRRT... RSA1314
0.0 -
-0.5-
-1.0
-1.5
LO E_XFC_NAT_u se
1.0-
0.5-
o.o H
-0.5-
-1.0-
-1.5-
-2.0 - -2.0 -
R S T
prk3RRT... RSA1314 prk3RRT...RSA1314 prk3RRT...RSA1314
146
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Appendix 8.A
NRSA 2008-09 & 2013-14 Expected Condition Models and
Condition Criteria
NOTES:
ฆ Uni-Fixed Models are fixed values of a metric that are uniform across all ecoregions and
both Boatable and Wadeable "Realms."
ฆ NULL MODELS are based on mean & SD for reference sites (prk3RRT_NRSA1314=R)
from NRSA0809 and NRSA1314.
ฆ Cond 1 and Cond lb MODELS are MLRs using reference sites (prk3RRT_NRSA1314=R)
from NRSA0809 and NRSA1314. Cond_l MLRs may have disturbance variable(s) as
predictors in cases where reference sites have anthropogenic disturbance that influences
response variable.
ฆ Cond ID MODELS are "All-Sites" MLRs using all sites (except Great Rivers) and
incorporate disturbance variables as predictors. We use 2-steps to calculate reference
expected values. First step is to calculate All-Sites Model Expected values then calculate
O/E values by setting disturbance to empirical minimum values for the ecoregion/realm.
Second value is to examine distribution of All-Sites Model O/E values within the Reference
Sites of the appropriate ecoregion/realm.
ฆ The expected reference value of the All-Sites Model OE is calculated from the reference site
distribution of All-Sites model O/E values (refOE mean & refSD) or a regression factoring
out disturbance in the reference sites (refOE y-intercept and refRMSE from disturbance
regression) *** note that there is no requirement that the disturbance variable be the same as
in the All-Sites model regression in fact it is likely to be a different variable because the
influence of the disturbance variable used in the "All-Sites Model" has already been
accounted for.
Condition Benchmarks for Riparian Human Disturbances
(RDist_COND) based on W1_HALL
We applied uniform condition benchmarks nationwide. The Low (1 Low), Medium (2 Medi), and
High (3 High) disturbance levels are analogous to the Good, Fair, Poor condition classification used
for the other indicators.
All Ecoregions and both Boatable and Wadeable sites
If Wl_Hall<0.33 then RDIST_COND= ฃ1 Low';
If Wl_Hall>=0.33 and Wl_Hall<1.5 then RDIST_COND='2 Medi';
If Wl_Hall> = 1.5 then RDIST_COND='3 High';
147
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Reference Condition Models for Channel Bed Sedimentation
based on Relative Bed Stability (LRBS use = LRBS_g08)
Coastal Plain (CPL) Boatable Sites
Cond_Null (eco9-B n=52)
RfNullMLRB S= -0.92405
RfNull SDLRB S= 1.33124
Coastal Plain (CPL) Wadeable Sites
Cond_l (eco9-W n=51):
LRBS_use= -1.67044 -0.77290(LXSlope_use)-0.49218(LXWidth_use) -0.12031(Wl_Hall)
R2=0.3497; AdjR2=0.3054; RMSE=0.73642; n=48/51; p=0.0003;pl<0.0001; p2=0.2637; p3=0.5799
Set Wl_Hall= 0 = minimum in ref sites:
RfEl_LRBS=-l.67044 -0.77290(LXSlope_use) -0.49218(LXWidth_use)
RfEl_RMSE_LRBS=0.73642
Northern Appalachian (NAP) Boatable Sites
Cond_Null (eco9-B n=47):
RfNullM_LRBS= -0.63226
RfNullSD_LRBS= 1.53888
Northern Appalachian (NAP) Wadeable Sites
Cond_l (eco9-W n=41):
LRBS_use= -0.64678 +0.32478(L_AreaWSkm2_use) -8.04380(W1_HAG)
R2= 0.2250; AdjR2=0.1842; RMSE=0.52529; n=41/41; p=0.0079;pl=0.0097;p2=0.1123
Set W 1_HAG=0 = minimum in ref sites:
RfEl_LRBS= -0.64678 +0.32478(L_AreaWSkm2_use)
RfEl_RMSE_LRBS=0.52529
Southern Appalachian (SAP) Boatable Sites
Cond_Null (eco9-B n=22):
RfNullMLRB S= 0.44138;
RfNullSD_LRBS=0.70357;
Southern Appalachian (SAP) Wadeable Sites
Cond_l (eco9-W n=32):
LRBS_use= -0.74349 +0.48842(L_AreaWSkm2_use) -0.75562(Wl_Hall)
R2= 0.2835; AdjR2=0.2341; RMSE=0.69081; n=32/32; p=0.0079;pl=0.0026;p2=0.0472
Set Wl_Hall=0 = minimum in ref sites:
RfEl_LRBS= -0.74349 +0.48842(L_AreaWSkm2_use)
RfEl RMSE LRBS=0.69081
148
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Northern Plains (NPL) Boatable Sites
CondlD (eco9-B n= 51)
"All-Sites Model" Regression on all 51 NPL boatable sites:
LRBS_use= -0.42002 +0.44371(L_AreaWSkm2_use)+1.26686(LXSlope_use) -
0.08698(AGws_X_KFct)
Set AGws_X_KFct = 0 = minimum for the ecoregion:
RfElD_LRBS= -0.42002 +(0.4437 l*L_AreaWSkm2_use) +(1.26686*LXSlope_use)
R2=0.5598; AdjR2=0.5311; RMSE All-Sites model= 0.61027; n=50/51;
p<0.0001;pl=0.0089;p2<0.0001;p3=0.0011
If don't have KFactor the following is very equivalent, as KFactors are close to 0.35:
LRBS use = -0.50236 +0.44371 (L_AreaWSkm2_use)+1.29164(LXSlope use) -
0.02628(PCT A G WS use)
RfOE 1DLRB S=LRB S use - RfElDLRBS
Regression using only NPL boatable Reference sites (n=28):
RfOElDLRBS = 0.15939 -0.02276(PCT_AG_WS_use)
R2=0.2322; AdjR2=0.2026; RMSE=0.51215; n=28/28; p=0.0094; pl=0.0094
~ Set PCT_AG_WS_use=0 = minimum in ref sites.
RfEOE 1 DLRB S= 0.15939
RfE_OElD_RMSE_LRBS=0.51215;
Northern Plains (NPL) Wadeable Sites
Cond lD (eco9-W n=314)
"All-Sites Model" Regression on all 314 NPL wadeable sites:
LRBS_use= -2.80718 +0.00084015(ELEV_PT_use) -0.70092(LXSlope_use) +0.64948(LXWidth_use)
-0.20932(W 1_HALL) -0.49739(WlH_Crop)
Set W1HALL and WI Crop = 0, the minima for the ecoregion:
RfElD_LRBS= -2.80718 +0.00084015(ELEV_PT_use) -0.70092(LXSlope_use)
+0.64948(LXWidth_use) ;
R2 = 0.3854; AdjR2=0.3754; RMSE All-Sites model =0.83720; n=314;
p<0.000 l;pl-3<0.0001;p4=0.0048;p5=0.0553;
RfOE 1 D LRBS=LRBS use - RfElD LRBS;
Regression using only NPL Wadeable Reference sites (n=52):
RfOElD LRBS = +0.19752 -0.31987(Wl_Hall);
R2=0.0280; AdjR2=0.0086; RMSE=0.83941; n=51/52; p=0.2356;pl=0.2356
RfE_OElD_LRBS= 0.19752
RfE_OElD_RMSE_LRBS= 0.83941
Southern Plains & Temperate Plains (SPL + TPL) Boatable Sites
Cond i (cenpl-B) n=47 ref sites from TPL, SPL, and NPL
LRBS_use= 1.44046 -0.32356*L_AreaWSkm2_use -0.02377*AG1KMCIRCLE
R2=0.1789; Adj R2=0.1416; RMSE=1.13936; n=47/47; p=0.0131;pl=0.0852;p2=0.0084
Set AG_lKMCircle=0 = minimum in reference sites:
RfEl_LRBS=l.44046 -0.32356(L_AreaWSkm2_use)
RfEl RMSE LRBS=1.14939
149
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Southern Plains (SPL) Wadeable Sites
Cond lD (eco9-W) n= 301
"All-Sites Model" Regression on all SPL wadeable sites:
LRBS_use= 0.89319 -0.06565(LAT_DD83) -0.09181(L_AreaWSkm2_use) -0.86897(LXSlope_use)
-0.24209(W1_HAG) -0.00308(AG_1KMCIRCLE) -0.02727(AGws_X_KFct)
R2=0.3525; Adj R2=0.3391; RMSE All-Sites=0.95158; n=297/301;
p<0.0001;pl=0.0002;p2=0.0519;p3<0.0001;p4=0.0155;p5=0.2490;p6=0.0049
Set W1HAG, AG lKMCircle, AGws x KFct = 0 = minima for SPL wadeable sites:
RfElD_LRBS= +0.89319 -0.06565(LAT_DD83) -0.09181(L_AreaWSkm2_use) -
0.86897(LXSlope_use)
RfOElD_LRBS= LRBSuse - RfElDLRBS
Regression on SPL wadeable ref sites:
RfElD_LRBS=
-0.00983 -0.83096(W1_HNOAG) -3.3658(Dam_dii) +0.6857(RdDen_ws_use) -
0.02242(PCT_Ag_ws_use)
R2 = 0.2616; Adj R2=0.1817; RMSE=0.99030; n=42/42;
p=0.0214;pl=0.0618;p2=0.2056;p3=0.0364;p4=0.0338
Set W1HNOAG, Dam_dii, PCT_Ag_ws_use = 0 = minima for ref sites;
Set RdDen_ws_use = 0 (Ref site mimimum = 0.19) zero leads to more lenient expected condition.
RfE_OElD_LRBS= -0.00983
RfE_OElD_RMSE_LRBS= 0.99030
Temperate Plains (TPL) Wadeable Sites
Cond_lD (eco9-W) ~ All-Sites Model Regression on all 344 TPL wadeable sites:
LRBS_use= 0.22205 +0.04387(LAT_DD83) +0.03596(LON_DD83) -0.49057(LXSlope_use)
-0.08247(W 1 _HAG) -0.01116(AG_ 1KMCIRCLE);
Set W1 HAG and AG_1 KMCIRCLE = 0 = minima for region:
RfElD_LRBS=0.22205 +0.04387(LAT_DD83) +0.03596(LON_DD83) -0.49057(LXSlope_use);
R2=0.1974; Adj R2=0.1854; RMSE-All-Sites =0.97639; n=342/344;
p<0.000 l;p 1=0.0556;p2=0.0074;p3<0.0001;p4=0.4971;p5<0.0001
RfOElD_LRBS= LRBS use - RfElD LRBS
Regression on TPL ref sites:
RfElD_LRBS= +0.21704 -0.83169(W1_HNOAG) +6.55336(WlH_Crop) -0.0228(Ag_lKmCircle)
-0.05988(PCT_Ag_ws_use) +0.19465(AgWS_x_KFct)
R2 = 0.3279; Adj R2=0.2633; RMSE=0.93335; n=58
/61 ; p=0.0007;p 1=0.0608;p2=0.0295;p3=0.0065;p4=0.0107;p5=0.0036
Set W1HNOAG, WlH_Crop, Ag_lKmCircle , PCT_Ag_ws_use, and AgWS_x_KFct = 0 = minima
for ref sites;
RfE_OElD_LRBS=0.21704 = y-intercept from above
RfE OE1D RMSE LRBS=0.93335
150
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Upper Midwest (UMW) Boatable Sites
Cond i (eco9-B n=36):
LRBS_use= 22.86206 -0.50298(LAT_DD83) -0.92704(W1_HALL)
R2=0.1820; Adj R2=0.1324; RMSE=1.25933; n=36/36; p=0.0363;pl=0.0113;p2=0.2028
Set W 1_HALL= 0 = minimum for regional ref sites
RfEl_LRBS=22.86206 -0.50298(LAT_DD83)
RfEl_RMSE_LRBS=l.25933;
Upper Midwest (UMW) Wadeable Sites
Cond_l (eco9-W n=43):
LRBSuse = -1.38974 -0.69289(LXSlope_use) -0.26824(W1_HALL)
R2=0.4103; Adj R2=0.3808; RMSE=0.92535; n=43/43; p<0.0001;pl<0.0001;p2=0.5347
Set W 1_HALL=0 = minimum for regional ref sites:
RfEl_LRBS= -1.38974 -0.69289(LXSlope_use)
RfEl_RMSE_LRBS=0.92535
Western Mountain fWMT) Boatable Sites
Cond_N (eco9-B n=43):
RfNullMLRB S= 0.36550
RfNullSD_LRBS=0.48996
Western Mountain fWMT) Wadeable Sites
Cond i (eco9-B n=69):
LRBS use = -0.77810 -0.31541(LXSlope_use) +0.48616(LXWidth_use)
R2= 0.2739; Adj R2=0.2516; RMSE=0.42995; n=68/69; p<0.0001;pl=0.0382;p2=0.022333
RfEl_LRBS=-0.77810 -0.31541(LXSlope_use) +0.48616(LXWidth_use)
RfEl RMSE LRBS=0.42995
Xeric (XER) Boatable Sites
Cond_N (eco9-B n=24):
RfNullMLRB S= 0.08641
RfNullSD_LRBS=0.98518
Xeric (XER) Wadeable Sites
Cond_l (eco9-W n=36):
LRBSuse = -2.01510 +1.33328(LXWidth_use)
R2=0.2333; Adj R2=0.2107; RMSE=0.79439; n=36/36 ; p=0.0028;pl=0.0028
RfEl_LRBS= -2.01510 +1.33328(LXWidth_use)
RfEl RMSE LRBS=0.79439
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
CONDITION ASSIGNMENTS FOR LRBS_use NULL MODELS:
RfNull25_LRBS=RfNullM_LRBS-(0.67*RfNullSD_LRBS);
RfNull05_LRBS=RfNullM_LRBS-( 1 65 *RfNullSD_LRBS);
RfOENullLRB S=LRB Suse -RfNullMLRB S;
LRB S_Cond_N='XXXX';
if LRBS_use<=RfNull05_LRBS then LRBS_Cond_N='Poor';
if LRBS_use>RfNull05_LRBS and LRBS_use<=RfNull25_LRBS
then LRBS_Cond_N='Medi';
if LRBS_use>RfNull25_LRBS then LRBS_Cond_N='Good';
If RfOENull_LRBS=. then LRBS COND N='XXXX';
If LRBS_use=. then LRBS_COND_N='XXXX';
CONDITION ASSIGNMENTS FOR LRBS_use COND 1 O/E MODELS:
RfOE 1 LRB S=LRBS_use-RfE 1 LRBS;
RfEl_25_LRBS=RfEl_LRBS-(0.67*RfEl_RMSE_LRBS);
RfE l_05_LRBS=RfE l_LRBS-( 1 65 *RfE 1RMSELRBS);
LRB S_Cond_ 1='XXXX';
if LRBS_use<=RfEl_05_LRBS then LRBS_Cond_l='Poor';
if LRBS_use>RfEl_05_LRBS and LRBS_use<=RfEl_25_LRBS
then LRBS_Cond_l='Medi';
if LRBS_use>RfEl_25_LRBS then LRBS_Cond_l='Good';
If RfEl_LRBS=. then LRBS_COND_l='XXXX';
If LRBS_use=. then LRBS_COND_l='XXXX';
CONDITION ASSIGNMENTS FOR LRBS_use COND ID ("All-Sites") O/E MODELS:
*** NOTE RfOE 1D LRBS=LRBS use-RfE 1D LRBS;
*** We base expectations on the distribution of OE in ref sites;
RfE_OElD_25_LRBS=RfE_OElD_LRBS-(0.67*RfE_OElD_RMSE_LRBS);
RfEOE 1D05LRB S=RfE_OE 1 D_LRBS-( 1.65* RfEOE 1DRMSELRBS);
if RfOElD_LRBS<=RfE_OElD_05_LRBS then LRBS_Cond_lD='Poor';
if RfOElD_LRBS> RfE_OElD_05_LRBS and RfOE 1D LRBS<=RfE_OE 1D 25 LRBS
then LRBS_Cond_lD='Medi';
if RfOElD_LRBS> RfE_OElD_25_LRBS then LRBS_Cond_lD='Good';
If RfOE ID LRBS=. then LRBS COND 1D='XXXX';
If LRBS_use=. then LRBS_C0ND_1D='XXXX';
152
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Reference Condition Models for Riparian Vegetation Cover
Condition
based on Logio(0.01+XCMGW)
Coastal Plain (CPL) Boatable Sites
Cond i (eco9-B n=52):
LPtO 1_XCMGW= 0.83657 +0.00658(LON_DD83) -0.06020(L_AreaWSkm2_use) -
0.57160(W1_HAG);
R2=0.2583; AdjR2=0.2121; RMSE=0.11862; n=52/52 ; p=0.0023; pl=0.0570; p2=0.0461; p3=0.0166
Set W1HAG = 0 = minimum for ref sites in region:
RfE 1 _LXCMGW=0.83657 +0.00658(LON_DD83) -0.06020(L_AreaWSkm2_use);
RfE 1RMSELXCMGW=0.11862;
Coastal Plain (CPL) Wadeable Sites
Cond_l (eco9-W n=51):
LPtO 1_XCMGW= -0.58185 -0.00700(LON_DD83)
R2= 0.0551; AdjR2= 0.0358; RMSE=0.15238; n=51/51 p=0.0972 pl=0.0972
RfE 1 LXCMGW = -0.58185 -0.00700(LON_DD83)
RfEl RMSE LXCMGW=0.15238
Northern Appalachian (NAP) Boatable Sites
Cond_l-eco9-B:
LptO 1 XCMGW = 2.51398 -0.05498(LAT_DD83)
-0.00786(AG_1KMCIRCLE) -0.79370(PCT_AG_WS_use) +2.68820(Agws_X_KFct)
R2 = 0.4025; AdjR2=0.3456;RMSE=0.15628; n=47/47 p=0.0002; pl=0.0056;
p2=0.0107;p3=0.0005;p4=0.0007
Set AG_1KMCIRCLE, PCT_AG_WS_use and AGws_X_KFct = 0 = minima for reference sites:
RfE 1_LXCMGW=2.51398 -0.05498(LAT_DD83)
RfEl_RMSE_LXCMGW=0.15628
Northern Appalachian (NAP) Wadeable Sites
Cond_l (eco9-W n=41):
LPtO 1_XCMGW=0.2114 l+0.09026(L_AreaWSkm2_use)
0.14456(W1_HALL)
R2 = 0.2411; AdjR2=0.1795; RMSE=0.12059; n= 41/41 p
Set W1 HALL = 0 = minimum for reference sites:
-0.30883(LXWidth_use) -
=0.0159;p 1=0.0894;p2=0.0130;p3=0.0293
RfE 1 _LXCMGW=0.21141 +0.09026(L_AreaWSkm2_use) -0.30883(LXWidth_use);
RfEl RMSE LXCMGW=0.12059
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Southern Appalachian (SAP) Boatable Sites
Cond_l (eco9-B n=22):
LPtO 1_XCMGW= 0.02698 -0.44778(W1_HAG)
R2 = 0.1689; AdjR2=0.1274;RMSE= 0.14138; n= 22/22 p=0.0574; pl=0.0574
Set W 1_HAG = 0 = minimum for reference sites:
RfE 1 LXCMGW = 0.02698
RfE 1 _RMSE_LXCMGW=0.14138;
Southern Appalachian (SAP) Wadeable Sites
Cond_l (eco9-W n=32):
LPt01_XCMGW= -0.14633+0.04120(L_AreaWSkm2_use) +0.00051106(ELEV_PT_use)
-0.16089(W 1 HALL);
R2=0.3232; AdjR2=0.2507; RMSE= 0.14090; n= 32/32 ; p=0.0111; pl=0.2142; p2=0.0028; p3=0.0429
Set W1HALL = 0 = minimum for reference sites:
RfE 1 LXCMGW = -0.14633 +0.04120(L_AreaWSkm2_use) +0.00051106(ELEV_PT_use);
RfEl_RMSE_LXCMGW=0.14090;
Northern Plains (NPL) & Southern Plains (SPL) Boatable Sites
Cond lD (CENPL-B n=249): ~ All-Sites Regression on All CENPL Boatable sites (NPL, SPL, & TPL)
LPtO 1_XCMGW= 1.80926 -0.02245(LAT_DD83) +0.01036(LON_DD83)
-0.24323(W1_HAG) -0.11970(RDDEN_WS_use)+0.00306(PCT_AG_WS_use)
R2 = 0.2485; AdjR2=0.2331; RMSE All-Sites= 0.36204; n= 249/249
p<0.0001;pl=0.0685;p2<0.0001;p3=0.0283;p4=0.0113.
Set W1HAG and PCT AG WS use = 0 = minimum in NPL+SPL ref sites
Set RDDEN_WS_use, = 0 (minimum for NPL+SPL ref sites = 0.043):
RfE 1 D_LXCMGW= 1.80926 -0.02245*LAT_DD83)+(0.01036*LON_DD83)
Regression on TPL CENPL ref sites:
RfOE 1DLXCMGW=LPt01 XCMGW - RfE ID LXCMGW;
RfEOE 1DLXCMGW = -0.08047 -0.01773(PCT_AG_WS_use)
R2 = 0.3141; AdjR2=0.2877; RMSE = 0.32423; n= 28/28; p=0.0019;pl=0.0019
Set PCT_AG_WS_use = 0 = min in ref sites
RfE OE 1 D LXCMGW = -0.08047 ;
RfEOE 1 D_RMSE_LXCMGW=0.32423;
154
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Northern Plains (NPL) Wadeable Sites
Cond lD (CENPL-W n=959) All-Sites Regression on All CENPL Wadeable sites (NPL, SPL, & TPL):
LPtO 1_XCMGW= 2.43249 -0.02325(LAT_DD83) +0.01579(LON_DD83)+0.16417(LXSlope_use)
-0.32696(W1_HAG) -0.00256(PCT_AG_WS_use)
R2 = 0.3126; AdjR2=0.3088; RMSE- All-Sites =0.48720; n=922/959; p<0.0001;pl-p4<0.0001;p5=0.0002
Set W1HAG and PCTAGWSuse = 0 = minima in ref sites of NPL (also SPL & TPL):
RfE 1DLXCMGW = 2.43249 -0.02325*LAT_DD83) +(0.01579*LON_DD83)
+(0.16417* LXSlope_use)
RfOE 1 D LXCMGW=LPt01 XCMGW - RfE 1 D LXCMGW;
Regression using only CENPL Wadeable Ref sites (155)
RfOE 1 D LXCMGW = -0.13159 -2.01216(Dam_dii) -0.02708(PCT_AG_WS_use) +
0.08125(AgWs_x_KFct);
R2 =0.1443; AdjR2=0.1270; RMSE=0.38555; n=152/155; p<0.0001;pl=0.0015;p2=0.0006;p3=0.0006
Set Damdii, PCT AG WS use & AgWs x KFct = 0 = minima in CENPL and NPL alone:
RfEOE 1 DLXCMGW = -0.13159;
RfE OE 1 D_RMSE_LXCMGW=0.38555;
Southern Plains (SPL) Boatable Sites
see combined NPL & SPL Boatable Sites above
Southern Plains (SPL) & Temperate Plains (TPL) Wadeable Sites
Cond_lb (spl-tpl-umw-W ~ ref sites only n=146):
LPtO 1_XCMGW= 1.25746 +0.01355(LON_DD83) -0.00024404(ELEV_PT_use)
-0.00636(AG_1 KMCIRCLE) -0.02587(PCT_AG_WS_use) +0.08055(AGws_X_KFct)
R2=0.3956; AdjR2=0.3735; RMSE=0.26692; n=143/146 p<0.0001;pl=0.0088;p2=0.0048;p3=0.0013;p4-
5<0.0001
Set AG1KMCIRCLE, PCT AG WS use & AGws X KFct = 0 = minima in both SPL and TPL ref
sites:
RfE 1 b LXCMGW = 1.25746 +0.01355(LON_DD83) -0.00024404(ELEV_PT_use);
RfElb_RMSE_LXCMGW=0.26692;
Temperate Plains (TPL) Boatable Sites
Cond_N (eco9-B n=22):
RfNullM_LXCMGW= -0.08249 ;
RfNullSD_LXCMGW= 0.15980 ;
Temperate Plains (TPL) Wadeable Sites
see combined SPL & TPL Wadeable sites above
155
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Upper Midwest (UMW) Boatable Sites
Cond lb (SPL + TPL+ UMW Boatable Ref sites n=55):
LPt01_XCMGW=l.52755 -0.03762(LAT_DD83) -0.33101(L_AreaWSkm2_use)
+0.17072(LXSlope_use) +0.82145(LXWidth_use)
R2-Square=0.3354 AdjR2=0.2822; RMSE=0.37273 n=55/55;
p=0.0003;pl=0.0057;p2=0.0019;p3=0.0284;p4=0.0003
Expected Ref condition model applied only to UMW Boatable sites:
RfE 1 b LXCMGW = 1.52755 -0.03762*LAT_DD83) -0.33101*L_AreaWSkm2_use)
+(0.17072*LXSlope_use) +(0.82145 *LXWidth_use)
RfE lb_RMSE_LXCMGW=0.3 7273
Upper Midwest (UMW) Wadeable Sites
Cond_l (eco9-W n=43):
LPtO 1_XCMGW= -0.13511 +0.05069(LXSlope_use)+0.17937(LXWidth_use)
-0.06747(W1_HALL)
R2=0.3303 AdjR2=0.2465; RMSE=0.12999 n=43/43 ; p=0.0028;pl=0.0115;p2=0.0025;p3=0.2867
Set W1HALL = 0 = minimum for ref sites:
RfEl_LXCMGW=-0.13511 +(0.05069*LXSlope_use) +(0.17937*LXWidth_use)
RfE 1RMSELXCMGW = 0.12999
Western Mountains (WMT) Boatable Sites
Cond_N (eco9-B n=43):
RfNullM_LXCMGW= -0.12272
RfNullSD_LXCMGW= 0.26191
Western Mountains (WMT) Wadeable Sites
Cond_l (eco9-W n=69):
LPtO 1_XCMGW=0.24290 -0.09638(L_AreaWSkm2_use) -0.00007192(ELEV_PT_use) -
. 11520(LXSlope_use)
R2= 0.2037; AdjR2=0.1669; RMSE= 0.15289; n=68/69; p=0.0019;pl=0.0063;p2=0.0024;p3=0.0425
RfE 1_LXCMGW= 0.24290 -0.09638(L_AreaWSkm2_use) -0.00007192(ELEV_PT_use) -
0.11520*LXSlope_use)
RfE 1 RMSELXCMGW=0.15289
Xeric (XER) Boatable Sites
Cond i (eco9-B n=24):
LPtO 1_XCMGW= -0.32820 +0.24638(W1_HNOAG) -0.15614(W1_HAG)
R2=0.2896; AdjR2=0.2220; RMSE=0.15263; n=24/24 ; p=0.0276;pl=0.0273;p2=0.1633;
Set W1HNOAG (positive beta) and W1HAG (negative beta) = 0 = minima for ref sites;
Note this results in lower ref mean, smaller RMSE, but lower (more lenient) percentile values than NULL
RfE 1 LXCMGW = -0.32820
RfEl RMSE LXCMGW=0.15263
156
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Xeric (XER) Wadeable Sites
Cond_l (eco9-W n=36):
LPtO 1_XCMGW= -0.21113 -0.19122(L_AreaWSkm2_use) +0.19148(LXSlope_use)
+0.65498(LXWidth_use)
R2=0.2294; AdjR2=0.1571; RMSE=0.25328; n=36/36; p=0.0374;pl=0.0695;p2=0.0730;p3=0.0086
RfE 1 LXCMGW = -0.21113 -0.19122*L_AreaWSkm2_use) +(0.19148*LXSlope_use)
+(0.65498*LXWidth_use);
RfEl RMSE LXCMGW=0.25328
CONDITION ASSIGNMENTS FOR RIPARIAN VEGETATION COVER NULL
MODELS:
RfNull25_LXCMGW=RfNullM_LXCMGW-(0.67*RfNullSD_LXCMGW);
RfNull05_LXCMGW=RfNullM_LXCMGW-(1.65*RfNullSD_LXCMGW);
RfOENull_LXCMGW=LPtO lXCMGW-RfNullMLXCMGW;
LXCMGW Cond N='XXXX';
if LPtO 1XCMGW<=RfNull05_LXCMGW then LXCMGW_Cond_N='Poor';
if LPtO l_XCMGW>RfNull05_LXCMGW and LPt01_XCMGW<=RfNull25_LXCMGW
then LXCMGW_Cond_N='Medi';
if LPtO l_XCMGW>RfNull25_LXCMGW then LXCMGW_Cond_N='Good';
if LPtO 1XCMGW =. then LXCMGW_Cond_N='XXXX';
CONDITION ASSIGNMENTS FOR RIPARIAN VEGETATION COVER COND 1 O/E
MODELS:
RfOE 1 LXCMGW=LPt01 XCMGW-RfE 1 LXCMGW;
RfE l_25_LXCMGW=RfE l_LXCMGW-(0.67*RfE 1RMSELXCMGW);
RfE l_05_LXCMGW=RfE l_LXCMGW-( 1.65 *RfE 1RMSELXCMGW);
LXCMGW_Cond_ 1 -XXXX';
if LPtO 1XCMGW<=RfEl_05_LXCMGW then LXCMGW_Cond_l='Poor';
if LPtO l_XCMGW>RfE 105 LXCMGW and LPt01_XCMGW<=RfEl_25_LXCMGW
then LXCMGW_Cond_l='Medi';
if LPtO l_XCMGW>RfE 125 LXCMGW then LXCMGW_Cond_l='Good';
If RfEl_LXCMGW=. then LXCMGW_Cond_l='XXXX';
if LPtO 1XCMGW =. then LXCMGW_Cond_l='XXXX';
CONDITION ASSIGNMENTS FOR RIPARIAN VEGETATION COVER COND lb O/E
MODELS:
RfE lb_25_LXCMGW=RfE lb_LXCMGW-(0.67*RfE IbRMSELXCMGW);
RfE lb_05_LXCMGW=RfE lb_LXCMGW-( 1.65 *RfE IbRMSELXCMGW);
LXCMGW_Cond_ 1 b='XXXX';
if LPtO 1XCMGW<=RfElb_05_LXCMGW then LXCMGW_Cond_lb='Poor';
if LPtO l_XCMGW>RfE lb_05_LXCMGW and LPt01_XCMGW<=RfElb_25_LXCMGW
then LXCMGW_Cond_lb='Medi';
157
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
if LPtO l_XCMGW>RfE lb_25_LXCMGW then LXCMGW_Cond_lb='Good';
If RfElb_LXCMGW=. then LXCMGW_Cond_lb='XXXX';
if LPtO 1XCMGW =. then LXCMGW_Cond_lb='XXXX';
CONDITION ASSIGNMENTS FOR RIPARIAN VEGETATION COVER COND ID
("All-Sites") O/E MODELS:
RfEOE 1D25LXCMGW=RfE_OE 1DLXCMGW-(0.67*RfE_OE 1DRMSELXCMGW);
RfEOE 1D05LXCMGW=RfE_OE 1 D LXCMGW-( 1.65* RfEOE 1 DRMSELXCMGW);
LXCMGW Cond 1D='XXXX';
if RfOE 1 D LXCMGW<=RfE_OE 1D 05 LXCMGW then LXCMGW_Cond_lD='Poor';
if RfOElD_LXCMGW> RfE_OElD_05_LXCMGW and
RfOE 1 D LXCMGW<=RfE_OE 1D 25 LXCMGW
then LXCMGW_Cond_ 1 D='Medi';
if RfOElD_LXCMGW> RfE_OElD_25_LXCMGW then LXCMGW_Cond_lD='Good';
If RfE_OElD_LXCMGW=. then LXCMGW_Cond_lD='XXXX';
if RfOE 1 D LXCMGW =. then LXCMGW_Cond_lD='XXXX';
158
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Reference Condition Models for Instream Fish Cover
based on Logio(0.01+XFC_NAT)
Coastal Plain (CPL) Boatable Sites
Cond_N (eco9-B n=52):
RfNullMLXFC NAT = -0.57048 ;
RfNullSD_LXFC_NAT= 0.23527 ;
Coastal Plain (CPL) Wadeable Sites
Cond_N (eco9-B n=51):
RfNullMLXF CNAT = -0.39218 ;
RfNullSD_LXFC_NAT=0.29820 ;
Northern Appalachian (NAP) Boatable Sites
Cond i (eco9-B n=47):
LPtO 1_XFC_NAT= -5.46962 -0.06654(LON_DD83) -0.46088(L_AreaWSkm2_use)
+0.92383(LXWidth_use)
-1.05887(WlH_Crop);
R2= 0.3404; AdjR2=0.2776; RMSE=0.31921; n=47;
p=0.0013;pl=0.0047;p2=0.0897;p3=0.0483;p4=0.0082
Set WlH Crop = 0 = minimum in ref sites:
RfE 1 _LXFC NAT = -5.46962 -0.06654(LON_DD83) -0.46088(L_AreaWSkm2_use)
+(0.92383(LXWidth_use);
RfE 1 RMSELXF C_NAT=0.31921;
Northern Appalachian (NAP) Wadeable Sites
Cond_l (eco9-W n=41):
LPtO 1_XFC_NAT= -0.08246 -0.26338(LXWidth_use) ;
R2=0.0736; AdjR2=0.0499; RMSE=0.28459; n=41; p=0.0862;pl=0.0862
RfE 1 _LXFC NAT = -0.08246 -0.26338(LXWidth_use);
RfEl_RMSE_LXFC_NAT=0.28459;
Southern Appalachian (SAP) Boatable Sites
Cond_l (eco9-W n=22) :
LPtO 1_XF C_NAT= -3.54570+0.07646(LAT_DD83) +0.22940(W1_HALL);
R2= 0.5343; AdjR2=0.4852; RMSE=0.17528; n=22/22; p=0.0007;pl=0.0089;p2=0.0065
Set W 1_HALL= 0 ~ note it is a positive association (mimimum in ref sites=0.03; in all sites=0):
RfEl_LXFC_NAT= -3.54570+(0.07646*LAT_DD83);
RfEl_RMSE_LXFC_NAT= 0.17528;
159
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Southern Appalachian (SAP) Wadeable Sites
Cond_l (eco9-W n=32):
LPtO 1_XFC_NAT= -2.89088 +0.06090(LAT_DD83) +0.0006263 l(ELEV_PT_use) -
7.37514(W1_HAG);
R2=0.4169; AdjR2=0.3544; RMSE=0.31006; n=32/32; p=0.0015;pl=0.0896;p2=0.0785;p3=0.0041
Set W 1_HAG = 0 = minimum for ref sites:
RfE 1 _LXFC NAT = -2.89088 +(0.06090*LAT_DD83) +(0.0006263 l*ELEV_PT_use);
RfEl_RMSE_LXFC_NAT= 0.31006;
CENPL (NPL, SPL, TPL) Northern, Southern & Temperate Plains Boatable Sites
Cond i (CENPL-B n=47):
LPtO 1 _XFC_NAT= 2.42961 -0.02335(LAT_DD83) +0.01564(LON_DD83) -
0.11096(L_AreaWSkm2_use)
-0.00934(AG_1KMCIRCLE) ;
R2=0.3446; AdjR2=0.2822; RMSE=0.32257; n=47/47;
p=0.0012;p 1=0.1204;p2=0.1228;p3=0.0400;p4=0.0070
Set AG_ 1KMCIRCLE = 0 = min for CENPL {Minima are 0%, 3.6%, 0.06% for NPL(n=33),
SPL(n=2), TPL(n=22)}:
RfE 1 _LXFC NAT = 2.42961 -0.02335*LAT_DD83)+(0.01564*LON_DD83) -
0.11096*L_AreaWSkm2_use);
RfE 1 RMSELXF CNAT = 0.32279;
CENPL (NPL, SPL, TPL) Northern, Southern & Temperate Plains Wadeable Sites
Cond i (CENPL-W n=155):
LPtO 1_XFC_NAT= -0.20615 +0.00409(LON_DD83) -
0.08735(L_AREAWSkm2_use)+0.00025270(ELEV_PT_use)
-0.0025 8(AG_ 1 KMCIRCLE) +0.043 32(URB_ 1 KMCIRCLE);
R2 =0.1740; AdjR2=0.1457; RMSE=0.33531; n=152/155;
p<0.0001;pl=0.5890;p2=0.0039;p3=0.0154;p4=0.1134;p5=0.0032
Set AG1KMCIRCLE and URB1KMCIRCLE =0 = minima for ref sites each of the 3 regions:
RfE 1 _LXFC NAT = -0.20615 +(0.00409*LON_DD83) -
0.08735*L_AREAWSkm2_use)+(0.00025270*ELEV_PT_use);
RfE 1 RMSELXF C_NAT=0.33531;
Upper Midwest (UMW) Boatable Sites
Cond i (eco9-B n=36)
LPtO 1 _XFC_NAT= 3.97716 +0.05232(LON_DD83) -0.20032(W1_HAG);
R2=0.2349 AdjR2=0.1885; RMSE=0.31606; n=36/36 ; p=0.0121;pl=0.0049;p2=0.8532
Set W1HAG = 0 = minimum in ref sites:
RfE 1 _LXF CNAT = 3.97716 +(0.05232*LON_DD83);
RfEl_RMSE_LXFC_NAT= 0.31606;
160
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Upper Midwest (UMW) Wadeable Sites
Cond_l (eco9-W n=43):
LPtO 1_XFC_NAT= -0.48451 +0.17605(L_AreaWSkm2_use) -0.35844(LXWidth_use) ;
R2 =0.0740; AdjR2=0.0277; RMSE=0.29010; n=43/43;p=0.2151;pl=0.0818;p2=0.1406
RfE 1 _LXFC NAT = -0.48451 +(0.17605 *L_AreaWSkm2_use) -0.35844*LXWidth_use);
RfEl_RMSE_LXFC_NAT= 0.29010;
Western Mountain fWMT) Boatable Sites
Cond i (eco9-B n=43):
LPtO 1_XF C_NAT= -1.40552 +0.48649(LXWidth_use)
-5.67454(WlH_Crop) -0.11975(RDDEN_WS_use) ;
R2=0.2408; AdjR2=0.1824; RMSE=0.23044; n=43/43; p=0.0124;pl=0.0175;p2=0.0077;p3=0.0654
Set WlH_Crop and RDDEN_WS_use = 0 = minima for ref sites:
RfE 1 _LXF CNAT = -1.40552 +(0.48649*LXWidth_use)
RfEl_RMSE_LXFC_NAT= 0.23044;
Western Mountain (WMT) Wadeable Sites
Cond_l (eco9-W n=69):
LPtO 1 _XFC_NAT= 1.57993 +0.01058(LAT_DD83) +0.01895(LON_DD83) -
0.08287(L_AreaWSkm2_use)
-11.24156(W1 HAG) -0.05374(RDDEN_WS_use);
R2 =0.3466; AdjR2=0.2939; RMSE=0.21669 n=68/69;
p<0.0001;pl=0.1652;p2=0.0013;p3=0.0414;p4=0.0054;p5=0.1064
Set W1_HAG and RDDEN_WS_use = 0 = minima for ref sites:
RfE 1 _LXFC NAT = 1.57993 +(0.01058*LAT_DD83) +(0.01895LON_DD83) -
0.08287(L_AreaWSkm2_use);
RfE 1_RMSE_LXFC_NAT=0.21669;
Xeric (XER) Boatable Sites
Cond i (eco9-B n=24):
LPtO 1_XFC_NAT= -0.03292 -0.00013276(ELEV_PT_use) -0.42159(LXWidth_use) ;
R2 =0.1266; AdjR2=0.1266; RMSE=0.31024; n=23/24; p=0.2582;pl=0.1323;p2=0.1973
RfE 1 _LXFC NAT = -0.03292 -0.00013276*ELEV_PT_use) -0.42159*LXWidth_use);
RfEl_RMSE_LXFC_NAT= 0.31024;
Xeric (XER) Wadeable Sites
Cond_l (eco9-W n=36):
LPtO 1 _XFC_NAT= 0.96284 +0.01132(LON_DD83)+0.18104(LXSlope_use) -19.86518(WlH_Crop);
R2=0.2738; AdjR2=0.2057; RMSE=0.24231 n=36/36; p=0.0155;pl=0.1628;p2=0.0431;p3=0.0353
note LXSlope distribution is similar across the range of the other model variables in all sites;
Set W lH_Crop = 0 = minimum in ref sites:
RfE 1 _LXF CNAT = 0.96284 +(0.01132*LON_DD83) +(0.18104*LXSlope_use);
RfEl_RMSE_LXFC_NAT=0.24231;
161
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
CONDITION ASSIGNMENTS FOR INSTREAM FISH COVER NULL MODELS:
RfNull25_LXF C_NAT=RfNullM_LXF C_NAT-(0.67* RfNullSD_LXFC_NAT);
RfNull05_LXF C_NAT=RfNullM_LXF C_NAT-(1.65 *RfNullSD_LXFC_NAT);
RfOENullLXF C_NAT=LPtO 1 _XF CNAT-RfNullMLXF CNAT;
LXFC NAT Cond N='XXXX';
if LPtO l_XFC_NAT<=RfNull05_LXFC_NAT then LXFC_NAT_Cond_N='Poor';
if LPtO l_XFC_NAT>RfNull05_LXFC_NAT and LPtO l_XFC_NAT<=RfNull25_LXFC_NAT
then LXFC_NAT_Cond_N='Medi';
if LPtO l_XFC_NAT>RfNull25_LXFC_NAT then LXFC_NAT_Cond_N='Good';
If LPtOlXFCNAT =. then LXFC_NAT_Cond_N='XXXX';
CONDITION ASSIGNMENTS FOR INSTREAM FISH COVER COND 1 O/E MODELS:
RfEl_25_LXFC_NAT=RfEl_LXFC_NAT-(0. 67* RfE 1RMSELXFCNAT);
RfE 1 05LXF CNAT=RfE 1 _LXF CNAT-(1.65* RfE 1 RMSELXF CNAT);
RfOE 1 _LXF CNAT=LPtO 1 _XF CNAT-RfE 1 _LXF CNAT;
if LPtO l_XFC_NAT<=RfE 105 LXFC NAT then LXFC_NAT_Cond_l='Poor';
if LPtO 1_XFC_NAT>RfE 105 LXFC NAT and LPt01_XFC_NAT<=RfEl_25_LXFC_NAT
then LXFC_NAT_Cond_l='Medi';
if LPtO l_XFC_NAT>RfEl_25_LXFC_NAT then LXFC_NAT_Cond_l='Good';
If RfEl LXFC NAT=. then LXFC NAT COND 1='XXXX';
If LPtOlXFCNAT =. then LXFC_NAT_Cond_l='XXXX';
162
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
9 Human Health Fish Tissue Indicator
Fish are time-integrating indicators of persistent pollutants, and contaminant bioaccumulation in fish
tissue has important human health implications. Contaminants in fish pose various health risks (e.g.,
cancer risks, and noncancer risks such as reproduction or neurological development impacts) to human
consumers. The NRSA 2013-14 human health fish tissue indicator provides information on the national
distribution of selected persistent, bioaccumulative, and toxic (PBT) chemical residues (e.g., mercury,
polychlorinated biphenyls (PCBs), and per- and polyfluoroalkyl substances (PFAS) in fish species that
people might catch and eat from rivers 5th order and greater in size in the conterminous United States.
Results of analyses of mercury, PCB, and PFAS fillet tissue concentrations are presented for this
indicator.
The human health fish tissue indicator field and analysis procedures described below were based on
EPA's National Study of Chemical Residues in Lake Fish Tissue (EPA 2009) and EPA's Guidance for
Assessing Chemical Contaminant Datafor Use in Fish Advisories, Volumes 1-2 (third edition) (EPA 2000).
9.1 Field Fish Collection
9.1.1 Fish Tissue Fillets
The NRSA 2013-14 crews collected fish for the fillet tissue indicator from a subset of rivers 5th order and
greater in size. The fish samples collected for fillet tissue analysis consisted of a composite of fish (i.e.,
five individuals of one target species)2 from each site. The fish had to be large enough to provide
sufficient tissue for analysis (i.e., 540 grams of fillets, collectively). Additional criteria for each fish
composite sample included fish that were:
All of the same species (for each site);
Harvestable size per legal requirements or of consumable size if there were no harvest limits; and
Similar size so that the smallest individual in the composite was no less than 75% of the total
length of the largest individual in the composite.
Crews were provided with a recommended list of target fish species (Table 9.1), but they could choose
an appropriate substitute if none of the recommended fish were available.
9.1.2 Fish Tissue Plugs
The NRSA 2013-14 crews collected fish for the tissue plug analysis from all river and stream sites
regardless of river or stream size. Two fish tissue plugs for mercury analysis were collected from two
fish of the same species (one plug per fish) from the target list. These fish are collected during the fish
2 Use of composite sampling for screening studies is a cost-effective way to estimate average contaminant concentrations while
also ensuring that there is sufficient fish tissue to analyze for all contaminants of concern. However, average concentrations
from composite samples may represent an over- or underestimation of a contaminant as compared to the actual concentration
in a single fish sample.
163
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
assemblage sample collection effort. A plug tissue sample was collected by inserting a biopsy punch into
a descaled thicker area of dorsal muscle section of a live fish. After collection, antibiotic salve was
placed over the wound and the fish was released.
Table 9.1 Recommended Target Species for Fish Tissue Indicator Sample Collection
Family Name
Common Name
Scientific Name
Length Guideline
(Estimated
Minimum)
Spotted bass
Micropterus punctulatus
-280 mm
Largemouth bass
Micropterus salmoides
-280 mm
Centrarchidae
Smallmouth bass
Micropterus dolomieu
-300 mm
Black crappie
Pomoxis nigromaculatus
-330 mm
White crappie
Pomoxis annularis
-330 mm
Channel catfish
Ictalurus punctatus
-300 mm
.34
'3
0>
Ictaluridae
Blue catfish
Ictalurus furcatus
-300 mm
s.
cn
Flathead catfish
Pvlodictis olivaris
-300 mm
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
benchmark represents the concentration that, if exceeded, may adversely impact human health. Fish
collection data were screened to exclude samples where non-target species were used or the average fish
length was less than 190 mm. Application of this benchmark to the fish tissue fillet composite data from
this study identifies the number and percentage of river miles in the sampled population containing fish
with mercury tissue fillet composite concentrations that are above the mercury human health fish tissue
benchmark. Results for the fish tissue fillet composite data are presented for the miles of 5th order and
larger rivers that could be sampled, and for the percentage of miles containing fish with tissue fillet
composite mercury concentrations that are above the benchmark. Results for the fish tissue plugs are
presented for all rivers and streams in the NRSA target population. To examine within-year variability,
analysts used the revisit sites to calculate a signal: noise estimate for the national mercury in fish tissue plug
dataset. The result was a S:N value of 6.35. Mercury concentration data from fish tissue plugs are available
to download from the NARS data webpage - https: / /www.epa.gov /national-aauatic-resource-
surveys/data-national-aquatic-resource-surveys. Mercury concentration data from fish fillet composite
samples are available to download from (https: / /www.epa.gov/fish-tech/2013-2014-national-rivers-
and-streams-assessment-fish-tissue-studv
9.3 PCB Analysis And Human Health Fish Tissue Benchmarks
A subset of 223 fish fillet composite samples collected at river sites were analyzed for PCBs. EPA
Method 1668C (EPA 2010) was used to analyze homogenized fillet tissue samples from each fish
composite sample. This method uses approximately 10 g of fillet tissue for analysis and provides results
for the full suite of 209 PCB congeners. The total PCB concentration for each sample was determined by
summing the results for any of the 209 congeners that were detected, using zero for any congeners that
were not detected in the sample. Two EPA human health fish tissue benchmarks were applied to
interpret total PCB concentrations in each fillet tissue composite sample: a benchmark of 18 ppb (wet
weight) for cancer risk and a benchmark of 73 ppb (wet weight) for noncancer risks {e.g., liver disease and
reproductive impacts). For more information on the human health fish tissue benchmarks, see Section
9.5. Application of these benchmarks to the total PCB fillet tissue composite data identifies the number
and percentage of river miles in the sampled population containing fish with total PCB concentrations
that are above each PCB human health fish tissue benchmark. Results are presented for the miles of
rivers, which are defined as 5th order or larger, that could be sampled and for the percentage of river miles
containing fish with fillet PCB concentrations that are above each total PCB human health fish tissue
benchmark. PCB concentration data from fish fillet composite samples are available to download from
(https: / / www.epa.gov/fish-tech/2013-2014-national-rivers-and-streams-assessment-fish-tissue-
9.4 PFAS Analysis And Human Health Fish Tissue Benchmark
Fillet tissue samples from 349 fish samples collected at river sites were analyzed for 13 per- and
polyfluoroalkyl substances (PFAS), including perfluorooctane sulfonate (PFOS), which is the most
commonly detected PFAS in fish. There are no standard EPA methods for PFAS analysis of tissue
165
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
samples, so the samples were analyzed by TestAmerica using a proprietary procedure developed by their
laboratory in West Sacramento, CA. That procedure, which utilizes approximately 1 g of fillet tissue for
analysis, uses high performance liquid chromatography with tandem mass spectrometry (HPLC-MS/MS)
and applies the technique known as isotope dilution to determine the concentration of each of the 13
PFAS. A human health fish tissue benchmark of 68 ppb (wet weight) was applied to interpret PFOS
concentrations in each fillet tissue composite sample. This benchmark was derived from the human
health reference dose published in EPA's Health Effects Support Document for PFOS in 2016 (EPA
2016)3. For more information on the human health fish tissue benchmarks, see Section 9.5. Application
of this benchmark to the PFOS fillet tissue data identifies the number and percentage of river miles in the
sampled population containing fish with fillet tissue PFOS concentrations that are above the PFOS
human health fish tissue benchmark. Results are presented for the miles of rivers (defined as 5th order or
larger) that could be sampled and for the percentage of river miles containing fish with fillet tissue PFOS
concentrations that are above the PFOS human health fish tissue benchmark. PFAS concentration data
from fish fillet tissue composite samples are available to download from (fattps://www.epa.gov/fish-
tt'ch /2013-2014-national-rivcrs-and-strcams-asscssmcnt-fish-tissuc-studv. Summary statistics,
including the number of detections for mercury, total PCBs, and each of the 13 PFAS are provided in
Table 9.2.
3 While there were other PFAS chemicals commonly detected in fish in this study, EPA does not have a reference dose to use in
development of a human health fish tissue benchmark for those other PFAS chemicals. Additionally, even though EPA has a
reference dose for PFOA, it was not commonly detected in fish in this study and therefore, EPA did not determine a
benchmark for PFOA in fish tissue.
166
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Table 9.2 NRSA 2013-14 Composite Fish Fillet Tissue Summary Data
Chemical
Number
Detection
MDLs
Measured
Weighted
Measured
of
Frequency
(PPb)
Minimum
Median
Maximum
Detections
(%)
Concentration
(PPb)*
Concentration
(ppb) *
Concentration
(ppb) *
Mercury
353
100
0.060
8.60
180
1070.00
Total PCBs
223
100
0.00006-
0.00098**
0.06
11.6
4616.59
Perfluorobutyric acid
(PFBA)
29
8
0.100
0.11
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
recommendation for methylmercury in 2001(EPA 2001b). For methylmercury, consistent with the 2008-
2009 NRSA report, EPA used this recommended fish tissue-based criterion as the benchmark to evaluate
mercury fish tissue results.5
PCBs: For PCBs, EPA has a recommended ambient water quality human health criterion that is expressed
as a concentration in ambient water, not as a fish tissue concentration. EPA used the chronic reference
dose (RfD) value,6 cancer slope factor, and equations found in EPA's Guidance for Assessing Chemical
Contaminant Data for Use in Fish Advisories (EPA 2000) to calculate fish tissue benchmarks for
evaluating the fish tissue results in this report. EPA developed two human health fish tissue benchmarks
for PCBs for the purpose of directly comparing to fish tissue results one based on carcinogenic effects
and one based on non-carcinogenic effects. Except for the revisions of equation inputs for body weight
and fish consumption rate described later in this section, the approach to develop the benchmarks for
PCBs was the same as EPA used in the 2008-09 NRSA.
PFAS: For the 2013-14 NRSA report, EPA tested fish tissue samples for 13 PFAS chemicals. EPA does
not have the toxicity information available to develop human health fish tissue benchmarks for most of
these PFAS chemicals. EPA has developed RfD values for PFOA and PFOS (EPA 2016b, EPA 2016c). In
the summers of 2013 and 2014, PFOA was only detected in 4% of fish fillet composite samples; thus, the
Agency did not develop human health fish tissue benchmarks for PFOA to evaluate results for this report.
However, during the 2013-14 sampling period PFOS was detected in 99% of fish fillet composite samples.
Therefore, EPA utilized the Agency's RfD value for PFOS and the equations found in EPA's Guidance
for Assessing Chemical Contaminant Data for Use in Fish Advisories to develop human health fish tissue
benchmarks to evaluate PFOS results for this report. For the 2008-09 NRSA, which was published prior
to development of the Agency's PFOS RfD, EPA used a human health fish tissue benchmark developed
by the State of Minnesota.
In using the equations found in its Guidance for Assessing Chemical Contaminant Data for Use in Fish
Advisories for developing human health fish tissue benchmarks for PCBs and PFOS, EPA revised two of
the inputs for use in the 2013-14 NRSA report compared to the 2008-09 NRSA report. EPA made this
change to ensure that the Agency's calculations most closely represent expected exposures and to increase
consistency between guidance for the fish advisory program and the water quality standards program.
Specifically, for both contaminants, EPA used updated body weights and fish consumption rates in the
equations for calculating the benchmarks. EPA used the body weight found in EPA's 2011 Exposure
Factors Handbook (EPA 2011) associated with the target population for which the RfD value or cancer
slope factor was developed for each contaminant.7
EPA previously used the average adult weight of 70 kg as recommended in EPA's Guidance for Assessing
Chemical Contaminant Data for Use in Fish Advisories. For the 2013-14 NRSA report, EPA used up-to-
date body weights: 80 kg for the average adult for the PCB benchmark and 75 kg for pregnant/lactating
women for the PFOS benchmark. EPA revised the default fish consumption rate in the equation to better
5 EPA notes that it analyzed the effect of changing equation inputs on the methylmercury benchmark and, even with updated
fish consumption rates and use of 75 kg body weight, the benchmark value is unchanged when rounded to appropriate
significant digits.
6 Chronic reference dose values represent the amount of a substance that a human can ingest each day without an
appreciable risk of negative health effects during a lifetime.
7 For PCBs, the reference dose value and cancer slope factor were based on non-developmental effects (immune, dermal and
ocular effects and cancer) so the target population is the general adult population. For PFOS, the reference dose value was
based on developmental effects, so the target population is pregnant/lactating women.
168
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
reflect the national average fish consumption rate as opposed to a value based on the U.S. Health and
Human Services Dietary Guidelines for Americans 2015-2020 (HHS 2015) which is a nutrition goal-based
recommendation. Previously for the NRSA 2008-09 report, EPA calculated the fish tissue benchmarks for
PCBs and PFOS based on the Dietary Guidelines of four, eight-ounce meals per month (29.8 grams/day).
In this NRSA 2013-14 report, EPA used the national default fish consumption rate of 22 grams/day from
EPA's Estimated Fish Consumption Rates for the U.S. Population and Selected Subpopulations (EPA
2014) that is used to calculate EPA's national ambient water quality human health criteria
recommendations. This revision is a better reflection of actual exposures. The difference in outcome using
the 2013-2014 versus 2008-2009 methodology is 40% versus 49% miles of river above the benchmark for
the PCB cancer benchmark; 17% versus 21% miles of rivers above the benchmark for the PCB non-
cancer benchmark; and 3% versus 8% miles of rivers above the benchmark for PFOS.
9.6 Literature Cited
HHS. 2015. 2015 2020 Dietary Guidelines for Americans. 8th Edition. U.S. Department of Health and
Human Services, Office of Disease Prevention and Health Promotion, Washington, D.C.
EPA. 2000. Guidance for Assessing Chemical Contaminant Data for Use in Fish Advisories, Volumes 1-2 (Third
Edition). EPA 823-B-00-007. U.S. Environmental Protection Agency, Office of Water,
Washington, DC.
EPA. 2001a. Appendix to Method 1631, Total Mercury in Tissue, Sludge, Sediment, and Soil by Acid
Digestion and BrCl Oxidation. EPA-821-R-01-013. January 2001. U.S. Environmental Protection
Agency, Office of Water, Washington, DC.
EPA. 2001b. Water Quality Criterion for the Protection of Human Health: Methylmercury. EPA-823- R-
01-001. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
EPA. 2002. Method 1631, Revision E: Mercury in Water by Oxidation, Purge and Trap, and Cold Vapor
Atomic Fluorescence Spectrometry. EPA-821-R-02-019. August 2002. U.S. Environmental
Protection Agency, Office of Water, Washington, DC.
EPA. 2009. The National Study of Chemical Residues in Fake Fish Tissue. EPA-823-R-09-006. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
EPA. 2010. Method 1668C, Chlorinated Biphenyl Congeners in Water, Soil, Sediment, Biosolids, and
Tissue by HRGC/HRMS, April 2010. U.S. Environmental Protection Agency, Office of Water,
Washington, DC.
EPA. 2011. Exposure Factors Handbook 2011 Edition (Final Report). EPA 600-R-09-052F. U.S.
Environmental Protection Agency, Office of Research and Development, Washington, D.C.
EPA. 2014. Estimated Fish Consumption Rates for the U.S. Population and Selected Subpopulations
(NHANES 2003-2010) Final Report. EPA 820-R-14-002. U.S. Environmental Protection Agency,
Office of Water, Washington, D.C.
EPA. 2016a. Health Effects Support Document for Perfluorooctane Sulfonate (PFOS). EPA 822-R-l 6-002. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
EPA. 2016b. Drinking Water Health Advisory for Perfluorooctanoic Acid (PFOA). EPA 822-R-16-005.
U.S. Environmental Protection Agency, Office of Water, Washington, DC.
EPA. 2016c. Drinking Water Health Advisory for Perfluorooctane Sulfonate (PFOS). EPA 822-R-16-004.
U.S. Environmental Protection Agency, Office of Water, Washington, DC.
169
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
IOEnterococci Indicator
The EPA has developed and validated a molecular testing method employing quantitative
polymerase chain reaction (qPCR) as a rapid approach for the detection of enterococci in
recreational water. NRSA used this method to estimate the presence and quantity of these fecal
indicator bacteria in the nation's rivers and streams. The statistical threshold value of 1,280
calibrator cell equivalents (CCE)/100 mL from EPA's 2012 Recreational Water Quality Criteria
document (RWQC) was then applied to the enterococci data to assess the recreational condition of
rivers and streams.
10.1 Field Collection
To collect enterococci samples, field crews took a water sample for the fecal indicator at the last
transect after all other sampling was completed. Using a pre-sterilized 250 mL bottle, they collected
the sample approximately 1 m off the bank at about 0.3 m (12 inches) below the water. Following
collection, crews placed the sample in a cooler and kept it on ice prior to filtration of two 50 mL
volumes. Samples were filtered and frozen on dry ice within 6 hours of collection. In addition to
collecting the sample, crews looked for signs of disturbance throughout the reach that would
contribute to the presence of fecal contamination to the waterbody.
10.2 Lab Methods
The sample collections and the laboratory method followed EPA's Enterococcus qPCR Method
1609.1 (USEPA 2015; available on-line at https://www.epa.gov/cwa-methods/other-clean-water-
act-test-methods-microbiological). As with EPA Draft Method A, used in the NRSA 2008-09
study, Method 1609.1 describes a quantitative polymerase chain reaction (qPCR) procedure for the
detection of DNA from enterococci bacteria in ambient water matrices based on the amplification
and detection of a specific region of the large subunit ribosomal RNA gene (IsrRNA, 23S rRNA)
from these organisms. Both methods use an arithmetic formula (the comparative cycle threshold
(Ct) method; Applied Biosystems, 1997) to calculate the ratio of enterococcus IsrRNA gene target
sequence copies (TSC) recovered in total DNA extracts from the water samples relative to those
recovered from similarly prepared extracts of calibrator samples containing a consistent, pre-
determined quantity of Enterococcus cells. Mean estimates of the absolute quantities of TSC
recovered from the calibrator sample extracts were then used to determine the quantities of TSC in
the water samples and then converted to CCE values as described in the section below. To
normalize results for potential differences in DNA recovery, monitor signal inhibition or
fluorescence quenching of the PCR analysis caused by a sample matrix component, or detect
possible technical error, Ct measurements of sample processing control (SPC) and internal
amplification control (IAC) target sequences were performed as described in Method 1609.1. The
primary differences between EPA Draft Method A (subsequently published as EPA Method 1611,
USEPA 2012a) and Method 1609.1 are that Method 1609.1 includes the IAC assay, an improved
170
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
polymerase reagent with greater resistance to inhibitory compounds and allows direct analyses of
undiluted sample DNA extracts. Analyses of diverse river water samples have indicated no
significant difference in the quantitative estimates obtained by the two methods (Sivaganesan et al.
2014).
10.3 Application Of Thresholds
10.3.1 Calibration
Estimates of absolute TSC recoveries from the calibrator samples were determined from standard
curves using EPA-developed plasmid DNA standards of known TSC concentrations as described in
Method 1609.1. Estimates of TSC recovered from the test samples were determined by the
comparative cycle threshold (Ct) method, as also described in Method 1609.1. Before applying the
EPA thresholds to the qPCR data, it was necessary to convert the TSC estimates to CCE values.
The standardized approach developed for this conversion is to assume 15 TSC/CCE (USEPA
2015). This approach allows the CCE values to be directly compared to the EPA RWQC values
(Haugland et al., 2014). A slightly modified approach was employed in the NRSA 2008-09 study to
obtain the same conversions of TSC to standardized CCE units.
10.3.2 Thresholds
For the data analysis of the enterococci measurements determined by Method 1609.1, EPA used
thresholds as defined and outlined in the 2012 RWQC document (USEPA 2012b). The document
contains the EPA's ambient water quality criteria recommendations for protecting human health in
marine and freshwaters. Enterococci CCE/100 mL values were compared to the EPA statistical
threshold value of 1280 CCE/100 mL8 (USEPA 2012b). Enterococci concentration data are
available to download from the NARS data webpage - https: / /www.epa.gov /national-aauatic-
resource-surveys/data-national-aquatic-resource-surveys.
To examine within-year variability, analysts used the revisit sites to analyze signal to noise of
enterococci concentrations and to analyze condition classes in a 2x2 contingency table. Condition
classes were defined as "above threshold" and "at or below threshold" based on the EPA
threshold value of 1,280 CCE/100 mL. The S:N ratio for concentration values was 0.36. However,
results from the contingency table analysis show that 78% of sites had the same condition class
during both visits (i.e., 64.7% of the 184 revisits that were assessed were at or below threshold in
both visits and 13.6% were above threshold in both visits) and 21.7% had mixed classes between
visits.
10.4 Literature Cited
8 Estimated Illness Rate (NGI): 32/1000 primary contact recreators. See USEPA 2012b for more information on additional
NGI statistical threshold values for the qPCR method.
171
-------
National Rivers and Streams Assessment 2013-2014 Technical Support Document
Applied Biosystems (1997) User Bulletin #2. ABI PRISM 7700 Sequence Detection System. Applied
Biosystems Corporation, Foster City, CA.
Haugland, R.A., S.D. Siefring, M. Varma, A.P. Dufour, K.P. Brenner, T.K. Wade, E. Sams, S. Cochran,
S. Braun, and M. Sivaganensan. 2014. Standardization of enterococci density estimates by EPA
qPCR methods and comparison of beach action value exceedances in river waters with culture
methods. Journal of Microbiological Methods 105, 59-66.
Sivaganesan, M., S. Siefring, M. Varma, and R.A. Haugland. 2014. Comparison of Enterococcus
quantitative polymerase chain reaction analysis results from midwestern U.S. river samples
using EPA Method 1611 and Method 1609 PCR reagents. Journal of Microbiological Methods 101:
9-17. Corrigendum 115, 166.
U.S. Environmental Protection Agency. 2012a. Method 1611: Enterococci in Water by TaqManฎ
Quantitative Polymerase Chain Reaction (qPCR) Assay. EPA-821-R-12-008. Office of Water,
Washington, DC.
U.S. Environmental Protection Agency. 2012b. Recreational Water Quality Criteria. EPA 820-F-12-
058. Washington, D.C.
U.S. Environmental Protection Agency. 2015. Method 1609.1: Unterococci in water by TaqManฎ
quantitative polymerase chain reaction (qPCR) assay with internal amplification control (IAC)
assay. EPA-820-R-15-009. US Environmental Protection Agency, Office of Water,
Washington, D.C.
172
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
11 Microcystes
Microcystin is a potent liver toxin produced by various cyanobacteria (blue-green algae).
Microcystins refers to the entire group of toxins, all of the different congeners, rather than just one
congener. Cyanobacteria can produce one or many different congeners at any one time, including
Microcystin-LR (used in the kit's calibration standards), Microcystin-LA, and Microcystin-RR. The
different letters on the end signify the chemical structure (each one is slightly different) which
makes each congener different.
11.1 Field Methods
Microcystin was collected as a grab water sample from Transect A (non-wadeable) or at the X-site
(wadeable) in a flowing portion near the middle of the channel. Water was collected using a 3 L
beaker and then transferred to a 500 mL bottle. The bottle was kept on ice and then stored frozen
until analysis.
11.2 Microcystin Analysis and Application of Benchmarks
Microcystin was measured using an enzyme-linked immunosorbent assay (ELISA) procedure with
an Abraxis' Microcystins-ADDA Test Kits. For freshwater samples, the procedure's reporting
range is 0.15 [J-g/L to 5.0 [J.g/L, although, theoretically, the procedure can detect, but not quantify,
microcystins concentrations as low as 0.10 [J.g/L. Microcystin concentrations were evaluated against
the EPA recommended swimming advisory level of 8 ng/L (USEPA 2019). Microcystin
concentration data are available to download from the NARS data webpage -
https:/ /www.epa.gov/ national-aquatic-resource-surveys/data-national-aquatic-resource-surveys.
To examine within-year variability, analysts used the revisit sites to calculate a S:N ratio estimate for
the national microcystin dataset. The result was a S:N value of 4.8. For this calculation, non-detect
values were excluded due to the fact that no variance between repeat sites when both were non-
detect may overestimate the S:N.
11.3 Literature Cited
Abraxis, "Microcystins-ADDA ELISA (Microtiter Plate)," Product 520011, R021412, Undated.
Retrieved January 2014 from
http://www.abraxiskits.com/uploads/products/docfiles/278 Microcystin%20PL%20ADD
A%20users%20R120214.pdf.
Abraxis, "Microcystin-ADDA ELISA Kit, Detailed Procedure," Undated. Retrieved January 2014
from http://www.abraxiskits.com/uploads /products/docfiles/253 PN520 .pdf.
173
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
James, R., et al, "Environmental Technology Verification Report: Abraxis Microcystin Test Kits:
ADDA ELISA Test Kit; DM ELISA Test Kit; Strip Test Kit," in Environmental
Technology Verification System Center 2010. Retrieved March 2013 from
http: / / nepis.epa.gov/Adobe/PDF/Pl POET, 6B.pdf
USEPA. 2P19. Recommended Human Health Recreational Ambient Water Quality Criteria or
Swimming Advisories for Microcystins and Cylindrospermopsin. EPA 822-R-19-PP1. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
174
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
12 From Analyses to Results
12.1 Extent And Risk Estimation And Assessment
A major goal of the National Aquatic Resource Surveys is to assess the relative importance of
stressors that impact aquatic biota on a national basis. EPA assesses the influence of stressors in
three ways: relative extent, relative risk, and attributable risk. The following discussion describes
the condition class assignments and calculations used in EPA's assessments. This discussion has
been adapted from a journal article by Van Sickle and Paulsen (2008).
12.1.1 Condition Classes
The NRSA database contains the field and laboratory data for all sampled sites, whether selected as
potential reference sites (i.e., hand-selected sites) or from the statistical design. For a number of
indicators, least-disturbed sites (i.e., reference sites described in Chapter 4) provided a benchmark
against which all other sites were compared and classified within each region. The condition classes
for each stressor and biological response were determined from data and observations from the
least-disturbed sites in each ecoregion and the continuous gradient of observed values at all sites.
For other indicators, for example microcystins, nationally established benchmarks were used.
The resulting three reference-based condition classes were defined as follows:
Good: As good as the best 75 percent of the reference sites;
Fair: Quality falling in between good and poor; and
Poor: Worse than 95 percent of the reference sites.
The resulting two condition classes based on the nationally established benchmarks were defined as
follows:
Above Benchmark
At or Below Benchmark.
The condition classes were then used to estimate the extent, relative extent, relative risk, and
attributable risk as described in the following sections. Only sites that were included in the
probability design and were evaluated as "Target_Sampled" were used to calculate extent and risk
estimates (i.e., SITETYPE = "PROB" and EVALSTAT = "TARGET_SAMPLED"). For sites that
were visited twice during NRSA 2013-14 (i.e., the -10% of sites that were used to asses indicator
S:N), only data from one site visit were included (i.e., "Visit 1"; in the datafiles these sites are
denoted as VISIT_NO = "1"). Sites that were hand-selected during 2013-14 (denoted as
SITETYPE = "HAND" and SITE_ID in the format of "XXRF-XXXX") were not included in
estimates.
When a nationally established thresholds was available for a given indicator (e.g., microcystins,
175
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
PFOA, fish tissue Mercury), the condition classification was set to either exceeding or not
exceeding. These indicators did not use reference condition to set the given benchmarks.
12.1.2 Estimating the Extent for Each Condition
The estimated extent E measures the prevalence of a particular condition k (good, fair, or poor).
For each Y, either a stressor or biological response, E provides an estimate of the miles of rivers
and streams in that condition. For example, E could be the estimated miles of rivers and streams
rated as "poor" for phosphorus concentrations in the lower 48 states.
The extent is estimated in two steps for each condition. The first step classifies each statistically
selected site into one of the three conditions for each Y. The second step estimates the miles
using the estimated survey weights Wi for each site i, classified into condition k. Applying weights
to the data allows inferences to be made about all river and streams in the target population, not
just the sites from which physical samples were collected. Each sampled site is assigned an
estimated weight for the number of miles that it represents. For example, one site might
represent 10,000 miles of rivers and streams in the entire target population, and thus, its sample
weight would be Wh; = 10,000. The following equation shows the estimation of extent (EXk) for
condition class k for each Y.
Eyk=IiWYk[ (12-1)
12.1.3 Relative Extent
For each particular Y (i.e., stressor or biological response), Relative Extent (REx) is the proportion of
"poor" miles in the target population. REx can also be interpreted as the probability that a river
or stream i chosen at random from the population will have poor conditions for Y. In statistical
terms where k=poor, this probability can be written as:
REYpoor = Pr0i = Poor) <12"2>
RE is estimated as the ratio of the sums of the sampling weights for the probability selected
sites / assessed as: (1) poor condition and (2) all sites regardless of condition. Where w^is the
number of sites in each condition, /?ฃ"can be expressed in statistical terms as follows:
vnpoor ^
fTrH Ypoor ^i=l wYpoori
REy = r = = nT~- n w (12"3)
Xpoor v,npoor ^ , ^"-fair ^ , 'Lqood ^ v '
wYpoori+Zi=1 wYfairi+^1 wYgoodi
176
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
12.1.4 Relative Risk
Relative risk (RR) measures the likelihood (that is, the "risk" or probability) of finding poor
(P) biological response B in a river or stream when the condition of a specific stressor S is also
poor. For relative risk, the good and fair sites are combined into a single non-poor (NP)
category. RR's likelihood is expressed relative to the likelihood of poor biological response
condition B in rivers and streams that have non-poor stressor conditions S. That is,
nn Pr (B=P\S=P)
RR = (12-4)
Pr(B=P)\S=NP) v y
To simplify the calculations, consider the notation in Table 12.1.
Table 12.1 Simplified Notation
Stressor (S)
Biological
Response (B)
Not-Poor (NP)
Poor (P)
Not-Poor
(NP)
Pr(B = NP\S = NP)= a
Pr(B = NP \ S = P) = b
Poor (P)
Pr(B = P \ S = NP) = c
Pr(B = P\S = P)= d
Using the simplified notation, RR is estimated as follows:
d
SH = (12-5)
a+c
RR =1.0 indicates "No association" between stressor and response, that is, poor biological
condition in a river or stream is equally likely to occur whether or not the stressor condition is
poor. RR < 1.0 indicates that poor response condition is actually less likely to occur when the
stressor is poor.
As a side note, using the simplified notation of Table 12.1, RESpoor from the previous section
(Equation 12-3) can be more simply written as:
n> b+d
RE c = (12-6)
Spoor a+b+c+d v )
for a stressor S in poor condition.
177
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
12.1.5 A ttributable Risk
Attributable risk (AR) estimates the change in ecological conditions when a stressor or biological
response is reduced or removed. AR is based on a scenario in which the stressor would be
restored through restoration activities to Not-Poor condition. For simplicity in terminology, this
discussion refers to the stressor as being "eliminated." AR is then defined as the proportional
decrease in the extent of poor biological response condition that would occur if the stressor were
eliminated from rivers and streams.
Attributable risk is derived by combining relative extent and relative risk from the proceeding
sections into a single estimate of the expected improvement in biological conditions if a
particular stressor is eliminated on a national or regional basis. Mathematically, AR is defined
as:
Pr(Y=P)-Pr (Y=P\S=NP)
AR = } / ' (12-7)
Pr (y=P)
We first calculated REy,estas shown in Equation 12-6 which is an estimate of Pr(Y = P). Then,
using the notation in Table 12.1,
REY.est c
ARest = (12-8)
RtY,est
Confidence intervals were calculated following the methodology described in Van Sickle
and Paulsen (2008).
12.2 Difference Analyses
One of the objectives of the NRSA is to track changes in the condition of rivers and streams over
time. Previously, EPA and partners reported on the condition of all rivers and streams in the NRSA
2008-09 and on the condition ofwadeable streams in the Wadeable Streams Assessment (WSA)
2004. The NRSA 2013-14 report presents the difference in percentage points of stream miles in
"good," "fair," and "poor" condition between NRSA 2008-09 and NRSA 2013-14. Additional
difference analyses were performed to determine the difference in the condition of the wadeable
streams population between 2004 and 2013-14 (see the NRSA interactive data dashboard). This
analysis does not represent a trend; until additional surveys are implemented, the NRSA can only
analyze differences between two survey time periods.
12.2.1 Data Preparation
Due to improvements in the sample frame for 2013-14, survey weights were updated for 2008-09 in
178
-------
National Rivers and Streams Assessment 2013-2014 Technical Report
order to make the population directly comparable to 2013-14. Analyses that were modified in NRSA
2013-14 (i.e., physical habitat metrics, fish MMI) were applied to previous applicable datasets (WSA,
NRSA 2008-09) in order for data to be directly comparable. Difference analysis was not conducted
for mercury in fish plugs or microcystin since these indicators were not included in NRSA 2008-09.
12.2.2 Analysis
Difference analysis was conducted through the use of the spsurvey 3.3 package in R (Kincaid and
Olsen, 2016). Within the GRTS (Generalized Random Tessellation Stratified) survey design,
difference analysis can be conducted on continuous or categorical variables. When using categorical
variables, difference is estimated by the difference in category estimates from the two surveys.
Category estimates were defined as the estimated proportion of values in each category (i.e., good,
fair, and poor categories). Difference between the two years was statistically significant when the
resulting error bars around the difference estimate did not cross zero.
12.3 Literature Cited
Kincaid, T.M., and A.R. Olsen. 2016. spsurvey: Spatial Survey Design and Analysis. R package
version 3.3.
Van Sickle, J., and S.G. Paulsen. 2008. Assessing the attributable risks, relative risks, and regional
extents of aquatic stressors. Journal of the North American Benthological Society 27:920-931.
179
------- |