Tennessee Integrated Assessment
of Watershed Health
A Report on the Status and Vulnerability of
Watershed Health in Tennessee
Prepared for—
US Environmental Protection
Agency Healthy Watersheds
Program
William Jefferson Clinton Building
1200 Pennsylvania Avenue, N.W.
Washington, DC 20460
Prepared by—
Kimberly Matthews, Michele Eddy,
and Phillip Jones (RTI)
Mark Southerland, Brenda Morgan,
and Ginny Rogers (Versar)
RTI International
3040 E. Cornwallis Road
Research Triangle Park, NC 27709
RTI Project Number 0213541.004.002.007
-------
Tennessee Integrated Assessment of Watershed Health
November 2015
EPA 841-R-15-002
Prepared by RTI International1 for the U.S. Environmental Protection Agency
Support for this project was provided by the EPA Healthy Watersheds Program
(www2.epa.gov/hwp)
Disclaimer
The information presented in this document is intended to support screening-level assessments of
watershed protection priorities and is based on modeled and aggregated data that may have been
collected or generated for other purposes. Results should be considered in that context and do not
supplant site-specific evidence of watershed health.
At times, this document refers to statutory and regulatory provisions, which contain legally binding
requirements. This document does not substitute for those provisions or regulations, nor is it a
regulation itself. Thus, it does not impose legally binding requirements on EPA, states, authorized tribes,
or the public and may not apply to a particular situation based on the circumstances.
Reference herein to any specific commercial products, process, or service by trade name, trademark,
manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation,
or favoring by the U.S. Government. The views and opinions of authors expressed herein do not
necessarily state or reflect those of the U.S. Government and shall not be used for advertising or
product endorsement purposes.
1 RTI International is a registered trademark and a trade name of Research Triangle Institute.
Tennessee Integrated Assessment of Watershed Health
1
-------
Acknowledgements
This document was prepared by RTI under contract with the U.S. Environmental Protection Agency
(EPA), Office of Water, Office of Wetlands, Oceans, and Watersheds. The following individuals are
acknowledged for their contributions to project planning, data acquisition, and review of draft materials:
• Nancy Arazan, U.S. Environmental Protection Agency
• Joy Broach, U.S. Army Corps of Engineers - Nashville District
• Rob Bullard, The Nature Conservancy
• Karina Bynum, Tennessee Department of Environment and Conservation
• Tim Diehl, U.S. Geological Survey
• Vivian Doyle, U.S. Environmental Protection Agency
• David Duhl, Tennessee Department of Environment and Conservation
• Pandy English, Tennessee Wildlife Resources Agency
• Veronica Fasselt, U.S. Environmental Protection Agency
• Joe Flotemersch, U.S. Environmental Protection Agency
• Jeff Fore, representing the West Tennessee River Basin Authority
• Laura Gabanski, U.S. Environmental Protection Agency
• Lisa Hair, U.S. Environmental Protection Agency
• Trisha Johnson, The Nature Conservancy
• Jeanette Jones, Tennessee Wildlife Resources Agency
• Susannah Kniazewycz, Tennessee Department of Transportation
• Rodney Knight, U.S. Geological Survey
• Brad Kreps, The Nature Conservancy
• Regan McGahen, Tennessee Department of Environment and Conservation
• Doug Norton, U.S. Environmental Protection Agency
• Shannon O'Quinn, Tennessee Valley Authority
• Sally Palmer, The Nature Conservancy
Tennessee Integrated Assessment of Watershed Health
2
-------
• Allison Reddington, U.S. Forest Service
• Matt Richards, Tennessee Department of Transportation
• Ben Rohrbach, U.S. Army Corps of Engineers - Nashville District
• Danny Sells, Tennessee Association of Conservation Districts
• Jimmy Smith, Tennessee Department of Environment and Conservation
• Sherry Wang, Tennessee Department of Environment and Conservation
• Joey Wisby, The Nature Conservancy
Tennessee Integrated Assessment of Watershed Health
-------
Table of Contents
Executive Summary
1. Introduction
1.1 Purpose and Intended Use
1.2 The Healthy Watersheds Program
1.3 Overview of Ecoregions in Tennessee
2. Methods Overview
2.1 Description of the Assessment Process
2.2 Conceptual Framework
2.3 Spatial Framework
2.4 Watershed Health Metrics
2.4.1 Landscape Condition Metrics
2.4.2 Geomorphic Condition Metrics
2.4.3 Hydrologic Condition Metrics
2.4.4 Water Quality Metrics
2.4.5 Habitat Condition Metrics
2.4.6 Biological Condition Metrics
2.5 Watershed Vulnerability Metrics
2.5.1 Land Use Vulnerability Metrics
2.5.2 Water Use Vulnerability Metrics
2.5.3 Climate Change Vulnerability Metrics
3. Results and Discussion
3.1 Watershed Health Index
3.2 Watershed Vulnerability Index
4. Assumptions and Limitations ¦
4.1 Spatial Framework ¦
4.2 Watershed Health Metrics and Sub-Indices ¦
4.2.1 Landscape Condition Sub-Index ¦
4.2.2 Geomorphic Condition Sub-Index ¦
4.2.3 Hydrologic Condition Sub-Index ¦
4.2.4 Habitat Condition Sub-Index <
4.2.5 Biological Condition Sub-Index <
4.3 Watershed Vulnerability Metrics and Indices ¦
5. Potential Applications of Assessment Results ¦
6. References ¦
Tennessee Integrated Assessment of Watershed Health
4
-------
Appendices
Appendix A Map Atlas
Appendix B Geomorphic Condition
Appendix C Hydrologic Condition
Appendix D Water Quality, Habitat and Biological Condtion Metric Modeling
Appendix E Data Analyses Methods and Correlation Results
Tennessee Integrated Assessment of Watershed Health
-------
List of Figures
ES-l. Watershed Health Index for Tennessee 10
1. Main river basins in Tennessee 13
2. Level III and IV ecoregions of Tennessee 14
3. Roadmap for this Assessment 18
4. EPA's six attributes of watershed health 19
5. Spatial framework for the Assessment 19
6. Difference between incremental and cumulative scales for quantifying landscape
variables for the same example catchment (dashed boundary) 20
7. Watershed health metrics used for the Assessment 21
8. Watershed vulnerability metrics used for the Assessment 33
9. Watershed Health Index and sub-index scores for Tennessee 38
10. Watershed Vulnerability Index and sub-index scores for Tennessee 39
List of Tables
1. Landscape and other variables used in statistical models 22
2. Classification of natural, semi-natural, and non-natural cover types 24
3. Variables used to determine Geomorphic Condition 26
4. Hydrologic metrics predicted using regression analyses (Knight et al., 2012) 28
5. Sample counts for filtered water quality parameters 30
Tennessee Integrated Assessment of Watershed Health
6
-------
Acronyms and Abbreviations
BRT
Boosted Regression Tree
CMIP5
Coupled Model Intercomparison Project 5
COMID
common identifier
CRAN
Comprehensive R Archive Network
EPA
U.S. Environmental Protection Agency
GCN
greatest conservation need
GFDLCM3
NOAA Geophysical Fluid Dynamics Laboratory Coupled Physical Model
GIS
geographical information system
GOF
goodness of fit
HAZ
hydrologically active zone
HUC
USGS hydrologic unit code
HWP
EPA Healthy Watersheds Program
IBI
Index of Biological Integrity
IQR
interquartile range
LCC
Appalachian Landscape Conservation Cooperative
MOU
Memorandum of Understanding
NHDPIus
National Hydrography Dataset Plus
NID
National Inventory of Dams
NLCD
National Land Cover Database
NOAA
National Oceanic and Atmospheric Administration
OSI
Open Space Institute
Q1
first quartile
Q3
third quartile
RBP
Rapid Bioassessment Protocol
SC
specific conductance
SE-GAP
Southeast Gap Analysis Program
SFC
streamflow characteristic
SSURGO
Soil Survey Geographic Database
STORET
EPA's Storage and Retrieval Data Warehouse
TDEC
Tennessee Department of Environment & Conservation
THWI
Tennessee Healthy Watershed Initiative
TMI
Tennessee Macroinvertebrate Index
TN
total nitrogen
TN SWAP
Tennessee State Wildlife Action Plan
TNC
The Nature Conservancy
Tennessee Integrated Assessment of Watershed Health
-------
TP
total phosphorus
TVA
Tennessee Valley Authority
TWRA
Tennessee Wildlife Resources Agency
USACE
U.S. Army Corps of Engineers
USGS
U.S. Geological Survey
WQ
water quality
Tennessee Integrated Assessment of Watershed Health
-------
Executive Summary
Tennessee's water resources are rich and varied, from the native brook trout streams of the Blue Ridge
Mountains in the east to the wide alluvial plains of the Mississippi River to the west. These resources are
a valuable asset to Tennessee and the protection and preservation of healthy waters in the state provide
recreational opportunities, clean drinking water, and other ecosystem services. This study was
conducted by the U.S. Environmental Protection Agency (EPA) in partnership with the Tennessee
Healthy Watershed Initiative (THWI). The THWI is a partnership among the Tennessee Department of
Environment and Conservation, the Tennessee Valley Authority, the Tennessee Chapter of The Nature
Conservancy, and the West Tennessee River Basin Authority working together to maintain and protect
water resources across the state by promoting communication, collaboration, and thoughtful planning.
The results from this study will be used to support the efforts of THWI and others working to protect
and restore the state's aquatic ecosystems.
The main goal of this Tennessee Integrated Assessment of Watershed Health (henceforth referred to as
the Assessment) is to identify healthy watersheds and characterize relative watershed health across the
state to guide future protection and restoration activities. A healthy watershed has the structure and
function in place to support healthy aquatic ecosystems. Key components of a healthy watershed
include:
• intact and functioning headwater streams, floodplains, riparian corridors, biotic refugia,
instream habitat, and biotic communities;
• a predominance of natural vegetation in the landscape; and
• expected hydrology, sediment transport, fluvial geomorphology, and disturbance
regimes for its location.
This report presents the methods and results of the Assessment and outlines proposed uses of the
results. The Assessment applied a system's approach that views watersheds and their aquatic
ecosystems as dynamic and interconnected systems in the landscape connected by surface and ground
water and natural vegetative corridors. Watershed health was quantified at the stream catchment scale
from existing geospatial datasets and from predictive models derived from data collected as part of
existing monitoring programs. This information was synthesized into several sub-indices that measured
aquatic ecological health and were combined into a comprehensive index of watershed health. The
potential for future degradation of watershed health was reported as a watershed vulnerability index.
An important facet of the Assessment is that it leverages existing efforts to analyze the characteristics of
watersheds and the aquatic ecosystems within them. Several agencies and organizations assess various
aspects of watershed health at statewide and regional scales. This project has used disparate datasets to
provide a more complete picture of watershed health across the state.
One output of the Assessment is a database of watershed health scoring metrics and catchment-based
information that can be used by THWI and other groups involved in watershed protection and
Tennessee Integrated Assessment of Watershed Health
9
-------
restoration planning. The database is intended to help identify healthy watersheds that are priorities for
local-scale assessment of protection opportunities. Several immediate uses of the database include
outreach and communication and prioritization of restoration and protection areas.
A second output is the integrated assessment framework developed by EPA and the THWI Technical
Team. This framework reflects our understanding of the interconnected nature of the physical,
chemical, and biological conditions of aquatic ecosystems; the significant effects of landscape- and
watershed-scale processes on aquatic ecosystem health; and the need to view water bodies as
connected parts within a larger system rather than as isolated units. The framework serves as a common
platform for the multistate agencies and organizations tasked with the protection and restoration of
Tennessee's water resources to collaborate and apply a unified approach rather than undertake
disjointed efforts. Over the long term, the existing framework can be updated as data gaps are filled and
improved assessment methodologies are identified.
The Assessment identifies relative health of watersheds across the entire state of Tennessee at the
catchment (approximately 1 square mile) level, based on metrics characterizing Landscape Condition,
Geomorphic Condition, Hydrologic Condition, Water Quality, Habitat Condition, and Biological
Condition. The scores from these six sub-indices were combined to create a Watershed Health Index.
The Vulnerability Index calculated from metrics characterizing potential threats to future watershed
health including Land Use, Water Use, and Climate.
Results can be presented for each metric, sub-index, or Watershed Health or Vulnerability Index at
multiple scales (i.e., catchment level or larger watersheds). Figure ES-1 illustrates the Watershed Health
Index at the catchment level. The highest scoring areas are in the Blue Ridge and Appalachian Mountains
in eastern Tennessee and scattered throughout the Interior Plateau in the central part of the state.
These areas are influenced by stable geomorphology, low deviation from natural streamflow, and
relatively good water quality and habitat conditions able to support diverse biological communities.
Figure ES-1. Watershed Health Index for Tennessee.
Tennessee Integrated Assessment of Watershed Health
10
-------
1. Introduction
1.1 Purpose and Intended Use
In 1996, the Tennessee Department of Environment and Conservation (TDEC) adopted a watershed-
based approach to monitoring and assessing their aquatic resources. This approach includes identifying
and prioritizing water quality challenges in the watershed, developing increased public involvement,
coordinating activities with other agencies, and measuring success through increased and more efficient
monitoring and other data gathering. Traditionally, these watershed efforts have focused on restoring
impaired streams, rivers, and lakes. Although some success has been achieved, many miles of stream
and acres of lake remain degraded, and new impairments continue to be identified. It is not only costly
to restore impaired water bodies, but also these water bodies are not able to provide the same
ecological, social, and recreation services as healthy aquatic ecosystems. Together, these issues call for
the expanded use of protection of healthy watersheds as a tool to preserve ecosystem services and
prevent the need for costly restoration.
The main goal of this Tennessee Integrated Assessment of Watershed Health (henceforth referred to as
the Assessment) was to characterize the relative health of watersheds in Tennessee to guide future
protection and restoration activities in the system. The Assessment synthesizes disparate datasets to
depict current landscape and aquatic ecosystem conditions throughout Tennessee. It is framed with the
recognition that the biological, chemical, and physical processes are interrelated and fundamentally
connected to the health of a water body and the maintenance of natural watershed processes. By
integrating information on multiple ecological attributes at several spatial and temporal scales, this
study provides a systems perspective on watershed health. This study was funded by the U.S.
Environmental Protection Agency's (EPA's) Healthy Watersheds Program and was performed in
conjunction with the Tennessee Healthy Watershed Initiative (THWI).
This report presents the methods, results, and intended applications of the Assessment. Readers are
asked to consider the following points regarding the scope of the Assessment as they review methods
and interpret results:
• The Assessment characterizes relative watershed health throughout Tennessee using a
collection of metrics that focus on the natural attributes of a watershed and its
freshwater ecosystems. No statement on the absolute condition of any watershed or
water body is made (e.g., attainment of designated uses), and results do not reflect the
influence of factors not considered for analysis.
• Data and information on relative watershed health are intended to support a screening-
level assessment of protection priorities across broad geographic areas (e.g., statewide
or within regional planning units). Assessment data should not supplant in-depth, site-
specific evidence of protection priorities, and conclusions drawn for smaller-sized areas
should be validated with site-specific information.
Tennessee Integrated Assessment of Watershed Health
11
-------
1.2 The Healthy Watersheds Program
EPA launched the Healthy Watersheds Program to support active protection of our nation's remaining
healthy watersheds (USEPA, 2012). A healthy watershed is one in which natural land cover supports
dynamic hydrologic and geomorphic processes within their natural range of variation, habitat of
sufficient size and connectivity to support native aquatic and riparian species, and physical and chemical
water quality conditions able to support healthy biological communities. Natural vegetative cover in the
landscape, including the riparian zone, helps maintain the natural flow regime and fluctuations in water
levels in lakes and wetlands. This, in turn, helps maintain natural geomorphic processes, such as
sediment storage and deposition that form the basis of aquatic habitats. Connectivity of aquatic and
riparian habitats in the longitudinal, lateral, vertical, and temporal dimensions helps ensure the flow of
chemical and physical materials and movement of biota among habitats.
Learn More Online:
Visit the EPA Healthy Watersheds Program Web site to view background material and project
reports: www2.epa.gov/hwp
EPA recommends using integrated assessments of watershed health to help states and others identify
healthy waters and prioritize candidate waters for protection and restoration. Integrated assessments
combine information on landscape condition, geomorphology, hydrology, habitat, water chemistry, and
biological communities. The Assessment synthesizes disparate datasets to depict current landscape and
aquatic ecosystem conditions throughout Tennessee. By combining multidisciplinary data from multiple
spatial scales, integrated assessments reflect our understanding of the:
• interconnected nature of the physical, chemical, and biological conditions of aquatic
ecosystems (lakes, rivers, streams, and wetlands);
• significance of landscape- and watershed-scale processes; and
• need to view water bodies as connected parts within a larger system rather than as
isolated units unaffected by their surrounding landscapes.
1.3 Overview of Ecoregions in Tennessee
Tennessee's water resources are rich and varied, from the native brook trout streams of the Blue Ridge
Mountains in the east to the wide alluvial plains of the Mississippi River to the west (Figure 1). Locally
high precipitation and diverse types of wetlands found especially in the eastern region of the state
provide habitat for many rare species of plants and animals. One small, shallow shoal within the Clinch
River is home to at least 35 mussel species, more than any other place on Earth. The Upper and Lower
sections of the Tennessee River sweep back and forth across the state for 360 miles before eventually
emptying into the Ohio River in neighboring Kentucky. The Cumberland River, located in north-central
Tennessee, flows into the state from the mountains of Kentucky through Nashville and back north into
Tennessee Integrated Assessment of Watershed Health
12
-------
Kentucky to join the Ohio River. The Mississippi River Basin dominates the western edge of the state.
Reelfoot Lake, in the northwestern corner of Tennessee, was created in the early 1800s by a series of
violent earthquakes and is now an important habitat for a large diversity of wintering and breeding
populations of waterfowl, including a significant population of wintering bald eagles.
Figure 1. Main river basins in Tennessee.
Eight Level III ecoregions made up of 25 smaller Level IV ecoregions have been delineated within
Tennessee (Figure 2). Below is a brief description of the Level III ecoregions and the subsequent Level IV
ecoregions with information adapted from Griffith and others (1997) and Omernik and Griffith (2009).
Tennessee Integrated Assessment of Watershed Health
13
-------
Figure 2. Level!!! and IV ecoregions of Tennessee.
Mississippi Alluvial Plain
Northern Mississippi Alluvial Plain
Mississippi Valley Loess Plains
] aiutf Hids
Loess Plains
Southeastern Plains
J Blackland Prairie
flatwoods/Alluvlal Prairie Margins
| | Southeastern Plains and Hilk
Fall Una Hills
Transition Hills
Interior Plateau
| | Western Pennyroyal Karat
Western Highland Rim
Eastern Highland Rim
Outer Nashville Basin
Inner Nashville Basin
Southwestern Appalachians
Cumberland Plateau
Sequatchi Valley
Plateau Escarpment
Central Appalachians
d Mountains
Ridge and Valley
Southern Limestone/Dolomite Valleys and tow Rolling Hills
I Southern Shale Valleys
I I Southern Sandstone Ridges
Southern Dissected Ridges and Knobs
Blue Ridge Mountains
I Southern Igneous Ridges and Mountains
Southern Sedimentary Ridges
H limestone Valleys and Coves
Southern Metasedimentary Mountains
• Mississippi Alluvial Plain: This riverine ecoregion along the Mississippi River is a flat,
broad floodplain dotted with river terraces and levees. The soils tend to be poorly
drained, and bottomland deciduous forest covered the region before most of the area
was cleared for agriculture. Within Tennessee, it comprises the Level IV Northern
Mississippi Alluvial Plain ecoregion entirely, which is bounded on the east by the Bluff
Hills and on the west by the Mississippi River. Most of this low-elevation region is
cultivated, with natural vegetation consisting of southern floodplain forest. Areas with
poor drainage may contain wooded swampland and oxbow lakes that serve as habitat
for waterfowl, raptors, and migratory songbirds, which are relatively abundant here.
• Mississippi Valley Loess Plains: This ecoregion in Tennessee abuts the Mississippi
Alluvial Plain and consists primarily of irregular plains with oak-hickory and oak-hickory-
pine natural vegetation. The primarily low-gradient streams in this region tend to have
silty substrates. The Bluff Hills and the Loess Plains are the two Level IV ecoregions
within this zone in Tennessee. Within the Bluff Hills ecoregion along the alluvial plain
boundary, smaller streams have areas of increased gradient and gravel substrate that
create aquatic habitats where unique, isolated fish assemblages more typical of upland
habitats can be found. The Loess Plains ecoregion is characterized by gently rolling,
irregular plains where most land has been cleared for agriculture, but some areas of
bottomland forest and cypress-gum swamp habitats remain. The region is crossed by
several large river systems with wide floodplains, where streams are murky with silt and
sand bottoms, and most have been channelized.
Tennessee Integrated Assessment of Watershed Health
14
-------
• Southeastern Plains: These irregular plains are located in the western half of the state,
just east of the Mississippi Valley Loess Plains, and are characterized by their higher
elevations and rolling topography. Streams in this area are relatively slow moving and
sandy bottomed. The majority of this ecoregion within Tennessee is delineated as the
Level IV Southeastern Plains and Hills ecoregion, where the natural vegetation type is
oak-hickory forest, grading into oak-hickory-pine to the south. The other four Level IV
ecoregions within the Southeastern Plains are located along the southern Tennessee
border and are very small in area within the state: Blackland Prairie, Flatwoods/Alluvial
Prairie Margins, Fall Line Hills, and Transition Hills. The Transition Hills exhibit the
highest elevations in the Southeastern Plains, and the streams resemble the sandy clear
streams of the Interior Plateau ecoregion directly to the east.
• Interior Plateau: The Interior Plateau is a diverse ecoregion consisting of five Level IV
ecoregions that stretch across a wide section of middle Tennessee. This ecoregion
contains the most diverse fish fauna in the state. The Level IV Western Highland Rim
ecoregion is characterized by rolling hills and streams with gravel and sand substrates
and relatively clear water. To the north, small sinkholes and depressions are common in
the Western Pennyroyal Karst ecoregion. The Inner and Outer Nashville Basin
ecoregions are in the center of the Interior Plateau and have distinctive fish fauna and
occasionally high densities offish because of productive, nutrient-rich streams. The
limestone cedar glades of the Inner Nashville Basin, a unique mixed grassland/forest
vegetation type with many endemic species, result in a distinct distribution of
amphibian and reptile species in this area. To the east, bordering the Cumberland
Plateau escarpment, the Eastern Highland Rim ecoregion contains numerous springs
and spring-associated fish fauna. Sinkholes and depressions are also common here
because of areas of karst terrain.
• Southwestern Appalachians: This ecoregion within Tennessee is characterized primarily
by the tablelands of the Level IV Cumberland Plateau ecoregion. These low mountain
areas receive slightly more precipitation with cooler annual temperatures than the
surrounding lower elevations. The eastern boundary of the ecoregion is relatively
smooth and notched by small stream drainages that flow eastward into the Great Valley
of East Tennessee (Ridge and Valley ecoregion). At the western boundary of the
Cumberland Plateau, the Plateau Escarpment Level IV ecoregion is characterized by
steep, forested slopes and fast-moving streams and waterfalls that have cut into the
limestone. The resulting ravines and gorges provide wet and cool environments that can
harbor distinct plant communities, such as hemlock stands along rocky streamsides and
river birch along floodplain terraces. A third Level IV ecoregion, the Sequatchie Valley,
outlines the Sequatchie River where erosion of broken rock to the south of the Crab
Orchard Mountains scooped out the long, narrow valley.
• Central Appalachians: This ecoregion in northern Tennessee is made up entirely of the
Cumberland Mountains Level IV ecoregion. The Cumberland Mountains are
characterized by rugged terrain, cool climate, and infertile soils that limit agriculture,
resulting in a mostly forested land cover. Steep slopes and narrow, winding valleys
separate mountain ridges. The natural vegetation is a mixed mesophytic forest,
although species diversity and abundance depend largely on microclimate. Coal mining
Tennessee Integrated Assessment of Watershed Health
-------
activities, including strip mining in the Cumberland Mountains, have caused siltation and
acidification of streams in this ecoregion.
• Ridge and Valley: Also known as the Great Valley of East Tennessee, this low-lying
region between the Blue Ridge Mountains to the east and the Cumberland Plateau on
the west is characterized by high aquatic habitat diversity and a diverse fish fauna.
Springs and caves are relatively numerous in this ecoregion. This region has four Level IV
ecoregions; the predominant are the Southern Limestone/Dolomite Valleys and the Low
Rolling Hills where landforms are mostly low rolling ridges and valleys. White oak
forests, bottomland oak forests, and sycamore-ash-elm riparian forests are the common
forest types. The Level IV Southern Shale Valleys ecoregion consists of lowlands and
rolling valleys, with well-drained soils that are often slightly acidic. Sandstone ridges and
valleys with sandy, poor soils typify the other two Level IV ecoregions (Southern
Sandstone Ridges and Southern Dissected Ridges and Knobs).
• Blue Ridge Mountains: The Blue Ridge Mountains of eastern Tennessee are
characterized by forested slopes and cool, fast-moving streams. Annual precipitation of
nearly 80 inches can occur on the well-exposed high peaks of the Great Smoky
Mountains that reach over 6,000 feet. The southern Blue Ridge is one of the richest
centers of biodiversity in the eastern United States. Blue Ridge streams have a distinct
fish fauna, with some containing brook trout, the only salmonid native to Tennessee.
Wetlands such as bogs, fens, and upland pools provide varying habitats among the
otherwise steep topography. These wetland communities, despite their very small
acreage, serve as important habitats for rare plant and animal species. Level IV
ecoregions within the Blue Ridge Mountains are the Southern Igneous Ridges and
Mountains, Southern Sedimentary Ridges, Limestone Valleys and Coves, and Southern
Metasedimentary Mountains.
Tennessee Integrated Assessment of Watershed Health
-------
2. Methods Overview
2.1 Description of the Assessment Process
This Assessment was conducted by EPA's Healthy Watersheds Program and in partnership with the
THWI. The THWI is a collaboration of federal, state, and nonprofit organizations committed to
maintaining and improving water resources in Tennessee watersheds. The THWI was launched under a
Memorandum of Understanding (MOU) executed by the Tennessee Department of Environment and
Conservation (TDEC), the Tennessee Valley Authority (TVA), the Tennessee Chapter of The Nature
Conservancy (TNC), and West Tennessee River Basin Authority in August 2011. The MOU signatories
recognize that many other governmental agencies and nongovernmental organizations have an interest
and a role in the health of Tennessee watersheds. The MOU and THWI Charter provide a structure that
others can participate in to the extent of their interest and ability, whether that is focused on a single
watershed, a region of the state, or the entire state (THWI, 2015).
Learn More Online:
Visit the Tennessee Healthy Watershed Initiative Web site to review more information about this
collaborative effort: https://www.tn.gov/environment/article/wr-ws-tennessee-healthv-watershed-
initiative
For this Assessment, TNC was the lead THWI member organization, and they assembled representatives
from federal and state agencies (e.g., TVA, TDEC, U.S. Geological Survey [USGS], U.S. Army Corps of
Engineers - Nashville District, Tennessee Wildlife Resources Agency [TWRA]) to serve on the Technical
Team. The Technical Team participated throughout the Assessment process by providing data and
information for the Assessment, reviewing the technical approach, and providing comments on the
preliminary analyses and draft report. Figure 3 illustrates the roadmap for the Assessment.
The first step of the Assessment was to create an inventory of available field monitoring and geospatial
data to assess current landscape, geomorphologic, hydrologic, habitat, water quality, and biologic
conditions throughout Tennessee. Data were gathered directly from the Technical Team and other
publically available sources such as EPA's Storage and Retrieval Data Warehouse (STORET) and USGS's
National Water Information System. Based on the available data, the technical approach for the
Assessment was prepared and reviewed by the Technical Team during an in-person meeting. The
meeting included a review of available data, discussion of the geospatial and statistical methodologies,
and discussion of the candidate watershed health and vulnerability metrics. Consensus on the key
technical aspects of the approach was achieved before implementing the technical approach. The
preliminary results were presented through a series of webinars to the Technical Team where the
technical approach was further refined.
Tennessee Integrated Assessment of Watershed Health
17
-------
Figure 3. Roadmap for this Assessment.
Select Indicators
Watershed Condition Watershed Vulnerability
~
Compute Indicator Values for all Watersheds with Available Data
Data Mining Compilation
~
Analyze Data
Statistical Modeling Geospatial Displays
~
Construct Multimetric Indices
Watershed Health Watershed Vulnerability
Communicate Results
Report and Database Workshop
Develop Strategic Priorities
Protection Restoration Monitoring
2.2 Conceptual Framework
EPA conceptualizes watershed health using six distinct but interrelated attributes: 1) Landscape
Condition, 2) Geomorphic Condition, 3) Hydrologic Condition, 4) Water Quality, 5) Habitat Condition,
and 6) Biological Condition (Figure 4; USEPA, 2012). An integrated watershed health assessment should
assess the condition of all six of these attributes using a variety of watershed health metrics.
Tennessee Integrated Assessment of Watershed Health
18
-------
Figure 4. EPA's six attributes of watershed health.
Landscape
Patterns of natural land cover, natural
disturbance regimes, lateral and longitudinal
> connectivity of the aquatic environment, and
continuity of landscape processes.
Geomorphic
Stream channels with natural
geomorphic dynamics.
| Habitat
Ij Aquatic, wetland, riparian, floodpfain,
lake, and shoreline habitat Hydrologic
connectivity.
Water Quality
Chemical and physical characteristics
of water.
4
Hydrologic
Hydrologic regime: Quantity and timing
y—of flow or water levels fluctuation. Highly
dependent on the natural flow (disturbance)
regime and hydrologic connectivity, including
surface-ground water interactions.
Biological
Biological community diversity, composition,
relative abundance, trophic structure,
condition, and sensitive species.
Data used to quantify watershed health metrics are selected to represent current conditions. Because
watershed health is a dynamic property that can vary with future changes in climate and human activity,
the Assessment also evaluates the vulnerability of watershed health to future conditions. Vulnerability is
quantified from a collection of watershed vulnerability metrics that characterize potential changes in
future land use, climate, and water use.
2.3 Spatial Framework
The spatial framework for conducting the Assessment was a network of small catchments represented
in the National Hydrography Dataset Plus (NHDPIus) Version 2. NHDPIus is a medium-resolution dataset
of all stream reaches in the nation and their corresponding catchments. Each NHDPIus catchment
represents the direct, or local, drainage area (median size of 0.6 square miles) for an individual stream
reach and has a common identifier (COMID) assigned to it in the dataset. A separate table identifies the
"from" and "to" COMID for every catchment in the dataset, giving a complete picture of the hydrologic
relationships between every catchment in the stream network at the 1:100,000 scale. Tennessee has
61,859 individual NHDPIus catchments (Figure 5).
Figure 5. Spatial framework for the Assessment.
Tennessee Integrated Assessment of Watershed Health
19
-------
The hydrologic relationships in NHDPIus allow for calculations of watershed characteristics (e.g.,
drainage area, stream length, land use) at both the incremental (within catchment boundaries) and
cumulative scales (within all upstream catchments) for any stream reach in Tennessee (Figure 6).
Cumulative values are included in the Assessment because of the potential for upstream conditions to
influence the health of a given stream reach. For example, high percent imperviousness in the
cumulative watershed is expected to influence downstream biological communities even though the
incremental imperviousness for the catchment may be low. In addition to its analytical benefits,
NHDPIus catchments can be aggregated to larger watershed scales. This allows for flexible reporting of
results at other watershed scales appropriate for multiple management or communication objectives.
Watershed health and vulnerability metrics were quantified on a catchment-by-catchment basis.
Calculating some metrics involved summarizing existing geospatial datasets to catchment-specific
values. Other metrics were quantified from modeled relationships between stream condition and
landscape variables. The NHDPIus dataset supports aggregation of incremental-to-cumulative data by
storing a unique numeric identifier for each catchment as well as upstream/downstream catchments.
Figure 6. Difference between incremental and cumulative scales for quantifying landscape variables
for the same example catchment (dashed boundary).
Channel Riparian Watershed
Incremental
Note: Variables quantified at the incremental scale summarize conditions within catchment boundaries only.
Variables quantified at the cumulative scale also summarize conditions throughout all upstream catchments,
expressed as a value of the downstream catchment.
A final note on the spatial framework of the Assessment relates to differences between the scale of
analysis and the intended scale of interpretation. Although NHDPIus catchments serve as analysis units,
results are not intended to be used to assess the condition of a single catchment. Rather, results should
Tennessee Integrated Assessment of Watershed Health
20
-------
be viewed over broad geographic areas to identify patterns and prioritize watersheds for in-depth, site-
specific assessments of protection needs. See Section 5 for more information on the potential uses of
the Assessment.
2.4 Watershed Health Metrics
Watershed health metrics were quantified on a catchment-by-catchment basis. Many metrics were
explored, but the selection of final metrics to use in the Assessment was determined by the robustness
of the dataset, the goodness of the model fit, the availability of data across the entire state, and input
from the THWI Technical Team.
A series of webinars was held with the THWI Technical Team to identify indicators of watershed health
that are most relevant to Tennessee and its stakeholders and for which data were readily available. The
discussion was framed around EPA's six attributes of watershed health to ensure that all aspects of
watershed health were explored. Ecological indicators were calculated for the following attributes:
1) Landscape Condition, 2) Geomorphic Condition, 3) Hydrologic Condition, 4) Water Quality, 5) Habitat
Condition, and 6) Biological Condition (Figure 7).
Figure 7. Watershed health metrics used for the Assessment.
Landscape Geomorphic Hydrologic Water Habitat Biologic
Condition Condition Condition Quality Condition Condition
Percentage
Natural Land
Cover
Erosive
Forces
Deviation in
Streamflow
Character-
istics1
Percentage
Natural Land
in HAZ
Resistive
Forces
Dam Storage
Ratio2
Stream Total
Nitrogen
Stream Total
Phosphorus
Stream
Specific
Conduct-
Rapid
Bioassess-
ment
Score
Macro-
invertebrate
IBI Rating
Habitat
Suitability
Score
Notes: 1 = unregulated streams, 2 = regulated streams; 3 = data only available for Ridge and Valley and Blue Ridge
ecoregions. HAZ = Hydrologically Active Zone; IBI = Index of Biological Integrity.
The methods used for this Assessment have been used in similar assessments for Wisconsin, California,
and Alabama. More information on these previous assessments is available on the EPA Healthy
Watersheds Program Web site (http://www2.epa.gov/hwp).
Tennessee Integrated Assessment of Watershed Health
21
-------
Three approaches were used to calculate metrics of watershed health for all catchments within
Tennessee. The first approach calculated metric values directly from geospatial data that have
representation across the entire state (e.g., land use, percentage of forest cover, and percentage of
imperviousness) and was used to calculate Landscape Condition and Geomorphic Condition Sub-Indices.
The second approach used geospatial data to represent catchment conditions (e.g., drainage area, and
soil conditions) as predictive variables in existing regression models to determine streamflow
characteristics. These data were used to determine the Hydrologic Condition Sub-Index. The third
approach used predictions from statistical models that relate landscape characteristics to stream
conditions. The statistical models were based on field-collected monitoring data throughout the state.
Because field-based monitoring data were not available for every catchment in the state, statistical
models were used to predict conditions in catchments without data. This approach was used to
determine Habitat Condition, Water Quality, and Biological Condition Sub-Indices. The combination of
actual and predicted data was used to rank the relative health of watershed conditions. The ranking of
the catchments is described in Appendix E.
The underlying sources of data for the Habitat Condition, Water Quality, and Biological Condition Sub-
Indices were field-based samples collected across the region through various state and federal
monitoring programs. Field-based data were not available for each of the NHDPIus catchments in
Tennessee. The existing monitoring data were used to predict habitat, water quality, and biological
condition in catchments without observed data using statistical regression models. These models
quantified relationships among landscape and other catchment characteristics and predict the values of
habitat, water quality, and biological condition for catchments without data. Landscape variables
described land cover, elevation, geology, and stream channel characteristics at both incremental
(catchment) and cumulative scales (see Figure 6). Other variables, such as sample date, corresponding
to field data were also used. Landscape and other variables quantified for statistical modeling are
presented in Table 1.
Table 1. Landscape and other variables used in statistical models.
Watershed Land
Cover
Percent natural lands, percent forest canopy, percent agriculture, percent disturbed, percent
forested land use, percent impervious surface (both within the catchment and cumulative)
Landscape
Catchment area, total drainage area, minimum and maximum stream elevation, mean catchment
elevation, mean soil erodibility (K factor), dominant Omernik III and IV ecoregion
Geology
Depth to bedrock, dominant surface geology, dominant bedrock type
Stream Channel
Characteristics
Sinuosity, stream length, stream order, channel slope
Sample
Sample date, sample month, sample year
Riparian Area Land
Cover
Percent natural lands, percent forested, percent agriculture, percent disturbed areas, percent land
use cover category, percent impervious surface (both within the catchment and cumulative)
Specific methods and statistical modeling approaches are described in the appropriate sections for each
Assessment component, and additional information is provided in Appendix B (Geomorphic Condition),
Appendix C (Hydrologic Condition), and Appendix D (Water Quality, Habitat, and Biological Condition).
Tennessee Integrated Assessment of Watershed Health
22
-------
Metrics of watershed health and vulnerability (see Section 2.5) were rank normalized for reporting the
metric, sub-index, and final index calculations. Rank normalization transforms one or more variables to a
uniform distribution and scale, typically from 0 to 100; this common scale allows for comparisons
between variables that may exhibit different units and scales. Rank normalization is also insensitive to
outlier or extreme values, which can overly compress a normalized distribution when other
normalization methodologies are applied (Mitchell, 2012). Once rank normalized on a common scale, a
correlation analysis between all possible pairings of the watershed health sub-indices was conducted to
determine whether there is any relationship between these calculated measures that would ultimately
prohibit combining the sub-indices into the Watershed Health Index without redundancy. The
correlation results supported the use of all sub-indices to create a multi-metric index representing
overall relative watershed health. More information on the rank-normalization methods and the results
from the Correlation Analysis is provided in Appendix E.
2.4.1 Landscape Condition Metrics
Landscape condition is described by the extent of natural land cover throughout a watershed and within
key functional zones such as floodplains, riparian areas, and wetlands. Land cover describes the physical
cover of the earth's surface, including natural and man-made vegetative cover and related land uses,
and plays a key role in the water cycle. When water falls as rain or melts as snow, the path the water
travels to reach streams, lakes, and rivers can either soak through the soil and become ground water or
travel over the land as runoff. The land cover determines which path the water takes; how long the
water needs to travel; and the amount of sediment, nutrients, and other constituents that are in the
water. For this Assessment, land cover representing naturally occurring communities such as forests and
wetlands is assumed to represent a landscape condition that does not negatively affect overall
watershed health. The first metric is based on the extent of natural land cover in the individual NHDPIus
catchment, and the second metric is based on the extent of natural land cover within the floodplains
and riparian areas of each catchment.
The 2011 National Land Cover Database (NLCD; Homer et al., 2015) was used to represent current
landscape conditions in Tennessee. The NLCD has a 15-class land cover classification scheme and a
spatial resolution of 30 meters. Additional NLCD products used elsewhere in this Assessment include the
percent developed imperviousness (Xian et al., 2011) and the percent forest canopy data products. The
NLCD 15-class scheme provides a coarse characterization of landscape conditions. To better represent
actual land cover conditions, the NLCD scheme was refined using two data sources: land cover mapping
from the Southeast Gap Analysis Program (SE-GAP) and mapping of managed forests produced by TNC
to create the 17-class scheme used in this Assessment (Table 2).
There are 71 SE-GAP classification units in Tennessee and each SE-GAP had a corresponding NLCD
category. This Assessment identified the nine naturally occurring SE-GAP communities that would
otherwise have been categorized as non-natural or semi-natural based on the NLCD classification. A new
land cover type listed as "SE-GAP" in Table 2 was the combination of these communities comprised of
rocky summits, cliffs, grass and shrub balds, and prairie lands. The other new land cover category,
Tennessee Integrated Assessment of Watershed Health
23
-------
Managed Forests, was used to differentiate forest lands that are managed and are often a monoculture
of one species. Management activities can alter natural hydrology and affect water quality, habitat, and
biological communities of surrounding streams. Therefore, this Assessment created a separate category
of managed forests as a semi-natural land cover. Two sources of data were used to identify managed
forests. One dataset was developed by the Open Space Institute (OSI; unpublished data) and is based on
a survey of land ownership. The second dataset was created by TNC (Barnett, 2015 unpublished data) as
a combination of SE-GAP communities classified as evergreen plantation and clear cut and merged with
county-level parcel data based on land ownership by timber companies.
The Landscape Condition metrics were based on the amount of natural, semi-natural, and non-natural
lands. Natural lands are defined as observed biological and physical condition of the Earth's surface that
represent lands without obvious human modification, including lands that have previously been
disturbed. Natural lands include forested lands, wetlands, cliffs, mountain balds, and prairie lands. Semi-
natural lands also have vegetation but are being maintained in a non-natural condition or are in the
process of recovering from disturbance. Semi-natural lands include shrublands, grasslands, and
industrial forests. Non-natural lands have been altered by human use and are actively maintained and
managed in way that is not consistent with natural vegetation composition. Non-natural lands includes
parks, lawns, cities, residential housing developments, row crops, and pasture lands.
Table 2. Classification of natural, semi-natural, and non-natural cover types.
HWP
Classification
NLCD Description and Classification Codes
Natural
Open water (11), deciduous forest (41), evergreen forest (42), mixed forest (43), woody wetlands (90),
emergent herbaceous wetlands (95), SE-GAP (new category)
Semi-natural
Shrub/scrub(52), grassland/herbaceous (71), managed forest (new category)
Non-natural
Developed, open space (21); developed, low intensity (22); developed, medium intensity (23);
developed, high intensity (24); barren land (31); hay/pasture (81); cultivated crops (82)
Percent Natural Land Cover: The significance of natural land cover to watershed health is represented in
the Assessment with the percent natural land cover metric. Percent natural land cover metric is
calculated as the sum of the area of natural cover types and 75% area of semi-natural cover type in a
catchment divided by the catchment's area and multiplied by 100.
Percent Natural Land Cover in Hydrologically Active Zone (HAZ): The proximity of land cover to
receiving rivers, streams, and lakes affects the degree to which the land cover will influence the
condition of that aquatic system. For the Assessment, this area is represented as the HAZ, which is a
combination of the riparian zone and the hydrologically connected zone developed by EPA Region IV as
part of the Watershed Index Online (EPA, 2014). The hydrologically connected zone is based on a
topographic index score and is contiguous to aquatic systems including streams and wetlands. The
riparian zone is calculated as a 100-meter-per-side buffer around the NHD flowlines. Percent intact HAZ
was determined for each catchment by combining the area of natural land cover types in the HAZ of
each catchment and 75% of the area of semi-natural lands in the HAZ of each catchment, dividing by the
Tennessee Integrated Assessment of Watershed Health
24
-------
HAZ area, and multiplying by 100. See Table 2 for a list of natural, semi-natural, and non-natural cover
types.
2.4.2 Geomorphic Condition Metrics
Fluvial geomorphology is the study of the shape of streams and their relationships with the landscapes
they flow through. Streams are dynamic systems, constantly carving and shaping their channel through
the movement of water. However, stream channels are subject to a wide variety of forces, both natural
and anthropogenic. The Geomorphic Condition describes how changes to the landscape affect stream
channel formation and evolution. It also helps predict whether a stream system can adjust to changes in
the watershed while maintaining its physical, biological, and chemical integrity. The principles of fluvial
geomorphology applied in this Assessment were developed by Leopold and others (1964) and Rosgen
(2006).
Geomorphic assessments are often completed to determine channel stability and resiliency to
watershed or reach-level disturbances. Channel stability does not mean that the stream's position and
form will remain fixed within the context of its landscape. Rather, streams in low-gradient, alluvial
valleys can naturally meander across a valley bottom, eroding an outside bend and depositing new
sediment on the inside of the bend. This form of lateral migration is generally a slow process and results
in only minor changes to a channel's dimensions (width, depth, area) even as the stream is actively
creating a new path across the terrain. This process is known as dynamic equilibrium. Channel resiliency
is the ability of the channel to maintain dynamic equilibrium as disturbances occur in the watershed or
along the stream corridor.
Streams often become unstable because of disturbances in the watershed that change the amount of
runoff and sediment that reaches the stream channel. Watershed and land use changes that cause
instability are called indirect disturbances. Streams can also become unstable because of direct changes
to the channel. Examples include channelization, removal of streamside vegetation, beaver dam and
wood removal, and in-stream mining. These direct and indirect disturbances can cause instability in the
vertical dimensions (e.g., streams can down-cut, becoming entrenched and isolated from their
floodplains), lateral dimension (e.g., destabilized channels may become unnaturally widened by erosion,
risking floodplain land loss while leaving a shallow stream that provides very poor habitat), or both.
Geomorphic stability is an important part of overall stream and watershed condition. Unstable channels
may increase fine sediment supply to the stream and downstream waterways, smothering benthic
habitats and eliminating the niche spaces where aquatic biota shelter from predators, lay eggs, and
forage for food. In addition, the subsequent increase in turbidity may lead to reduced primary
productivity, increasing stress throughout the food web, and lead to changes in water chemistry (Castro
et al. 1995). Other consequences of instability may include threats to human infrastructure and a
reduction in natural flood controls.
The evaluation of Geomorphic Condition was based on multiple watershed variables to determine the
balance of erosive and resistive forces at work within a catchment. Little field-based data were available
Tennessee Integrated Assessment of Watershed Health
25
-------
to characterize geomorphic condition of the watersheds; therefore, geospatial data were used as
proxies for field-based measurements (Table 3). For this Assessment, Geomorphic Condition for each
catchment was characterized in three ways: erosive forces and susceptibility to erosion, resistive forces
that abate erosion, and the integration of erosive and resistive forces to gauge the overall potential for
geomorphic stability.
Table 3. Variables used to determine Geomorphic Condition.
Variable
Source (and Method)
Selection Rationale
Annual Flow
Calculated mean annual runoff (eastern Tennessee)
and mean summer streamflow (western
Tennessee)1
Erosive factor - Approximation of the strength of
the hydrologic regime that drives channel formation
and change
K-Factor
Soil erosion potential, attribute in SSURGO
Erosive factor - Natural susceptibility of soils along
the stream channel to erosion
Land Cover
(Impervious-
ness)
NLCD 2011, Impervious Surface
Erosive factor - Representative of anthropogenic
influence within a catchment that will lead to
changes in the timing, volume, and velocity of runoff
entering a stream channel
Depth to
Bedrock
Average depth to bedrock along each flowline was
calculated using the Generalized Geologic Map of
the Conterminous United States (Nicholson et al.,
2005)
Resistive factor - Representative of the limit of
change possible within a stream channel (i.e., a
restrictive layer that is not reformed by erosion)
Land Cover
(Forest,
Impervious)
NLCD 2011, Cumulative NLCD, NLCD 2011 Canopy
Resistive factors - Approximation of the natural
control and infiltration of runoff in a catchment
Land Cover
(Natural
Land in the
HAZ)
NLCD 2011, NLCD 2011 Canopy
Resistive factor - Used as a gauge for an
undisturbed riparian, with vegetative cover that
provides natural checks on channel migration and
widening
1 These variables were calculated as metrics of Hydrologic Condition and are described in Section 2.5.3.
Erosion Metric: Three factors were used to assess the potential for the stream to incise (lower its bed)
and to erode laterally and cause channel widening: percent impervious cover (for the cumulative land
area draining to the catchment), soil erodibility, and annual flow (i.e., hydrologic force at work in the
stream channel). Impervious cover increases the amount of water reaching the stream channel by runoff
and that increase in runoff can lead to channel erosion. The erosion potential for the soil is measured as
K-factor and was obtained from the Soil Survey Geographic Database (SSURGO) developed by the
Natural Resources Conservation Service. Soils having a high silt content are the most erodible of all soils.
They are easily detached and tend to crust and produce high rates of runoff. Values of K for these soils
tend to be greater than 0.4. Medium-textured soils, such as the silt loam soils, have moderate K values,
about 0.25 to 0.4, because they are moderately susceptible to detachment and they produce moderate
runoff. Coarse-textured soils, such as sandy soils, have low K values, about 0.05 to 0.2, because of low
runoff even though these soils are easily detached. Soils high in clay have low K values, about 0.05 to
0.15, because they are resistant to detachment. K-factor classifications were based on research by Jones
and others (1996). The streamflow is a calculated quantitative measure of the stream's ability to do
work, typically defined as moving sediment.
Tennessee Integrated Assessment of Watershed Health
26
-------
Resistance Metric: A high percentage of impervious cover in a catchment and an increase in stream
power (i.e., an increase in the erosive forces at work in a stream channel) do not mean that the
streambed will incise or that the stream channel will widen or migrate at an accelerated rate because
other factors can limit erosion. The surrogates used to represent these resistive factors include depth to
bedrock or other constraining layers such as hard claypan, percent forest cover (for the cumulative land
area draining to the catchment), and the percent of natural cover within the HAZ. Streambeds that are
composed of bedrock will not incise, regardless of changes in hydrology (runoff). Bedrock is a major
form of grade control for the streambed. As the depth to bedrock increases, the potential for stream
incision also increases. Streambeds that have a claypan restrictive layer, although not as resistant to
erosion as bedrock, will also exhibit less incision and will erode more slowly. The percent of forest cover
also mitigates the potential for incision by lowering the volume of runoff from the watershed (opposite
to percent impervious cover). Vegetation within the HAZ may anchor stream banks and constrain
excessive meandering. Vegetation with deep roots, especially near the channel, holds the bank together,
thereby reducing the potential for erosion and subsequent stream migration.
A simple continuous scoring model (range: 0-100 points) was used for each factor included in the
analysis (see Appendix B for details of the analysis, including the variable, their values, and scoring
system was applied to those values). Each factor was scored so that higher point values indicated the
factor would have a positive effect on stream resilience and stability (e.g., a lower percentage of
impervious cover is less likely to alter the natural flow regime in a catchment and therefore would score
more points than a catchment with a higher percentage of impervious cover). The three factor scores
were then averaged to produce an erosion and a resistance metric score, respectively. If a value for a
particular factor was not available, this factor was dropped from the average. In this manner, each
catchment was given an erosion metric score and a resistance metric score. These two metric scores
were averaged to determine the Geomorphic Condition Sub-Index.
2.4.3 Hydrologic Condition Metrics
A stream's flow regime refers to its characteristic pattern of flow magnitude, timing, frequency,
duration, and rate of change (Poff et al., 1997). The flow regime plays a central role in shaping aquatic
ecosystems and the health of biological communities. Alteration of natural flow regimes (e.g., more
frequent floods) can reduce the quantity and quality of aquatic habitat, degrade aquatic life, and result
in the loss of ecosystem services. Therefore, to assess Hydrologic Condition, we used metrics related to
the flow regime in unregulated streams to determine which segments most closely resemble the natural
flow regime through reference watersheds and were therefore assumed to be healthy. In regulated
systems (i.e., streams below large dams), we used the ratio of the storage behind the dams to the
expected mean annual natural streamflow to determine which regulated segments have lower volumes
of storage compared with streamflow and therefore had greater potential to influence the natural flow
regime. Individual dam operations and rules have the potential to mitigate these influences; however,
these factors are not included in this Assessment. Information on dam operations is provided in
Appendix C for qualitative assessment.
Tennessee Integrated Assessment of Watershed Health
27
-------
Unregulated Streams/Catchments
The regional regression models developed by Knight and others (2012) were used for the Tennessee and
Cumberland River basins to determine streamflow characteristics (SFCs) by catchment for the eastern
portion of the state. In the western portion of the state, SFC regression models developed by Law and
others (2009) were used. Both models rely on basin characteristics such as drainage area and underlying
geologic and soil conditions to predict the SFCs. Each calculated SFC at the catchment level was
compared with a range of values expected for streams under natural or reference conditions, which was
determined by first selecting catchments exhibiting more natural land cover and then calculating the
interquartile range (IQR) of calculated SFC values for those catchments. The absolute value of the
deviation from this range (giving both high and low deviations equal weight and assuming any deviation
from natural is impactful to the flow regime), if any, was calculated for each SFC. The summation of
deviations for all SFCs by catchment provided the overall hydrologic condition metric for unregulated
streams (see Appendix C for more details).
East Tennessee: The USGS regional regression model (Knight et al., 2012) consists of 19 separate
regression equations that predict a single SFC for unregulated streams (Table 4). The regression
equations were derived based on 231 USGS streamflow monitoring sites (drainage areas spanning 1.67
to 3,035 square miles) across the two basins using geospatially derived sub-basin characteristics as
independent variables. An additional USGS study (Knight et al., 2014) related SFCs to fish community
structure and found that eight SFCs were influential to fish species richness in each of the three
ecological regions (i.e., Blue Ridge, Ridge and Valley, and Interior Plateau) covered by the Tennessee
River basin. Significant SFCs identified in that study were recalculated for this Assessment and used as a
starting point to select a subset of ecologically relevant metrics to use in the eastern portion of the state
to evaluate the streamflow regime (Appendix C).
Table 4. Hydrologic metrics predicted using regression analyses (Knight et al., 2012).
Hydrologic
Characteristic
Metrics
Magnitude
Mean annual runoff (MA41), maximum October streamflow (AMH10), streamflow value exceeded 85% of
time (e85), median September daily flow (Sep_med), rate of streamflow recession (LRA7)
Ratio
Average 30-day maximum (LDH13), base flow (ML20), constancy (TA1), number of day rises (RA5)
Frequency
Frequency of moderate flooding (three times the median annual flow [FH6]) and (seven times the median
annual flow [LFH7])
Variability
Variability of March streamflow (MA26), variability in base flow (LML18), variability of annual minimum daily
average streamflow (LDL6), variability in high-pulse duration (LDH16), variability in low-pulse count (FL2)
Date
Annual minimum flow (TL1), annual maximum flow (TH1), flow direction reversals (RA8)
Bold metrics are those found to be influential to fish species richness in the Tennessee River basin by Knight et al. (2014).
To calculate the SFCs for each NHDPIus catchment, the independent variables were calculated for each
catchment through new geospatial analyses. A final selection of SFCs was made after comparing values
for each SFC at the 231 monitoring site locations between the USGS study and this Assessment to
determine which SFCs diverged least from the original study and best represented the gauged flows.
Ultimately, three SFCs were chosen to provide a measure of Hydrologic Condition in eastern Tennessee:
Tennessee Integrated Assessment of Watershed Health
28
-------
mean annual runoff (MA41), date of annual minimum flow (TL1), and variation in high-pulse duration
(LDH16). These three metrics assess magnitude, timing, and variability in the flow regime and, therefore,
provide information on multiple aspects of the flow regime. A reference range for values of each of
these three metrics was calculated by ecoregion by designating catchments with forest land use in the
upper quartile of the range of values as reference catchments. The IQR for SFC values for these
reference catchments designates the reference range from which deviations for all other catchments
were calculated. The total deviation among the three SFCs for each catchment becomes the Hydrologic
Condition metric for the eastern portion of the state.
West Tennessee: Assessment in the western portion of the state relied on the same independent
variables calculated from geospatial analyses but applied regression models of various SFCs developed
by Law and others (2009) from 124 streamflow gauges within the western portion of the state for flow
magnitudes, frequencies, and durations. After comparison to the original data and consideration of the
general hydrologic conditions of the western basins (i.e., groundwater driven), three SFCs were chosen
for the western portion of Tennessee: lowest consecutive 7-day average flow that occurs every 10 years
(7Q10), mean-summer streamflow in June through August (MS), and daily mean streamflow exceeded
10% of the time (qlO). The regression equations determined by Law and others (2009) and used for this
Assessment estimate flow magnitudes. To make relative comparisons across the region and apply the
reference region deviation method, resulting SFCs were normalized by drainage area of the catchment.
As in East Tennessee, the reference range was developed by selecting catchments with forest land use in
the upper quartile range of values as reference catchments. Then the IQR for each of the three SFC
values was used to designate the reference range and the total deviation from this range among the
three SFCs became the Hydrologic Condition metric for the western portion of the state.
Regulated Streams
Dams have a major impact on natural riverine hydrology, primarily through changes in the timing,
magnitude, and frequency of low and high flows. Major dams within Tennessee were identified as
having greater than 10,000 acre-feet of normal storage. All catchments within the state that are located
at or downstream of a dam of this size were classified as being regulated and were assessed separately
from the unregulated portions of streams within the state.
Dam Storage Ratio: The ratio of the volume of water impounded by dams and the average annual
predevelopment streamflow serves as an indicator of potential hydrologic alteration. Using data from
the Tennessee State Wildlife Action Plan (TN SWAP), dams with greater than 10,000 acre-feet of normal
storage were identified throughout the state and upstream areas. From this selection, any dams with a
primary purpose of recreation were removed to eliminate counting natural lakes with spillways from the
Assessment (located in western Tennessee). Catchments downstream from each of these dams were
identified and indexed to all upstream dams. The storage volume of all upstream reservoirs was
summed for each of these identified catchments based on the values provided by the TN SWAP dataset.
Natural streamflows were provided for each catchment by the NHDPIus dataset. The dam storage ratio
was calculated as the storage volume divided by the expected natural streamflow with conversion to
Tennessee Integrated Assessment of Watershed Health
29
-------
number of days. Ratio values near or below zero indicate there is little dam storage (0 to less than 1 day
of storage volume compared with annual average streamflow) within the drainage area; therefore, the
hydrology is closer to natural conditions. Ratio values near one indicate drainage areas where the
volume of water stored behind dams is approximately equal to the volume of water flowing through the
catchment on a daily average across the year. Values greater than one highlight basins dominated by
dams and reservoir storage over streamflow. Because the three different metrics were assessed as rank
normalized values, they were combined into a single statewide coverage for comparison of all
hydrologic regions of the state.
2.4.4 Water Quality Metrics
Water quality refers to a suite of physical and chemical parameters present in surface and ground
waters. Water quality parameter values are influenced by a complex set of factors that interact across
multiple spatial and temporal scales. Parameter values in a healthy watershed should fall within the
range of naturally occurring variation for that water body. Values that exceed this natural variation can
negatively impact the physical, chemical, and biological processes that occur in surface waters; these
changes can in turn alter the fundamental dynamics of aquatic ecosystems.
The Water Quality assessment primarily considers "naturally occurring parameters," a phrase that refers
to physical and chemical characteristics that are likely to be present in surface waters regardless of
watershed health. To assess Water Quality, a relational database was created using data collected by
TDEC. Based on a survey of available data, feedback from the THWI Technical Team, and the goals of the
Assessment, three stream water quality metrics were selected for analysis:
• stream total nitrogen concentration,
• stream total phosphorus concentration, and
• stream specific conductance.
These parameters were characterized as the annual post-2000 median value for catchments with five or
more unique samples. Appendix D lists additional filters that were applied to the water quality
parameter values. The final number of catchments with data are listed in Table 5.
Table 5. Sample counts for filtered water quality parameters.
Water Quality Parameter
Catchments with Data
Total Nitrogen
1,690
Total Phosphorus
1,828
Specific Conductance
1,677
These parameters were selected by the Technical Team because they represent important aspects of
water quality health in Tennessee and monitoring data of sufficient spatial and temporal resolution was
available to produce a relative statewide ranking of water quality condition on the catchment level.
Tennessee Integrated Assessment of Watershed Health
30
-------
Additional physical and chemical parameters may also impact water quality health in a given catchment.
For example, parameters such as pH or organic and inorganic contaminants may be influential at local or
regional scales. The interpretation of the Water Quality Sub-Index should always consider local
conditions.
Not all catchments in the state have monitoring data, so statistical modeling was used to relate
observed water quality data with watershed-scale predictor variables. Watershed-predictor variables
were summarized at multiple spatial scales and evaluated land cover, geology, impervious cover, and
other influential physical factors that were used as proxy variables for watershed-scale processes such
as runoff, buffering capacity, and other fate and transport mechanisms (see Table 1). The modeled
relationships between the selected water quality parameters and predictor variables were then used to
predict water quality values for catchments without monitoring data. For this Assessment, a tree-based
modeling approach called boosted regression tree modeling was used to characterize and predict water
quality condition for all NHDPIus catchments in the state. Detailed information on the statistical
modeling and water quality results can be found in Appendix D.
2.4.5 Habitat Condition Metrics
Aquatic habitat is an essential component of watershed health because it is often the limiting factor for
biological communities. Even where water quality is in good condition, biota may not attain reference
condition without the physical habitat features of their natural environment. Indeed, habitat loss and
degradation are usually cited as the primary factors affecting biological diversity in streams worldwide.
Habitat degradation can result from a variety of human impacts occurring within the water body itself or
in the surrounding watershed. Typical in-stream impacts include sedimentation, channelization, and
bank erosion and filling, such as mountain top removal. Urban development, timber harvesting,
agriculture, livestock grazing, energy extraction, streamflow barriers/impediments (hydrologic
alteration), and the draining or filling of wetlands are well-known examples of human activities affecting
stream habitat at the watershed scale.
Rapid Bioassessment Protocol (RBP) Score: The RBP is a commonly used tool for assessing the condition
of physical habitat in streams. RBP data are usually collected whenever benthic macroinvertebrate or
fish samples are taken in streams. These data include the presence and quality of stream banks, riffles,
pools, and other physical features that provide habitat for aquatic species. The RBP index has associated
condition classes (usually four to five categories on a 20-point scale) that are benchmarked to
conceptual reference conditions. TDEC usually assesses RBP statewide, wherever biological assessments
are performed. Their database contains 4,175 unique sampling sites sampled from 2000 to 2013,
distributed among the Level III ecoregions.
Habitat Suitability Scores: TNC worked with TWRAto develop TN SWAP (TNC, 2012). Toward this end, a
GIS and relational database management system was developed to manage the large amounts of data
on species of greatest conservation need (GCN), their habitats, and problems affecting these species and
habitats. Terrestrial, aquatic, and subterranean habitats were classified and mapped, and habitat
Tennessee Integrated Assessment of Watershed Health
31
-------
preferences for over 600 faunal species were assigned by taxonomic experts involved in the planning
effort. The database contains over 52,000 occurrences of 664 GCN species, and 15,878 of these records
are for aquatic species. Additionally, the database contains over 131,000 records of host fish for mollusk
species occurrences. Predictor variables derived from the NHDPIus dataset were used to map expected
species distribution for known occurrences, based on habitat availability. A model was developed to
rank all catchments from low GCN species priority to very high. This model was based on an index of
relative viability for every species occurrence in the database. Additionally, stream segments upstream
and downstream of those with known occurrences were evaluated and scored, based on species
rarity/legal status score, viability score, flow distance, and a percent deviation of mean annual flow
volume from that of the stream segment with the documented occurrence. Dams were not crossed
when assessing potential occurrence extent. Once an overall habitat priority for each stream segment
was calculated, the amount of habitat represented by each stream segment was used to weight the
overall catchment score.
2.4.6 Biological Condition Metrics
Biological Condition is the most integrative of the six healthy watershed attributes, representing the
cumulative effect of biogeochemical features of the environment (including historical factors) on the
communities of organisms within the watershed ecosystem. This may include affects from features that
are unknown or impossible to measure. The use of biological condition indices (such as Indices of
Biological Integrity [IBIs]) depends critically on the definition of reference condition so that naturally
depauperate areas are not viewed as degraded. Benthic macroinvertebrates and fish assemblage
metrics were used for this Assessment.
Benthic Macroinvertebrate Score: TDEC uses two different benthic macroinvertebrate sampling
methods and index formulations (TDEC, 2011) to assess biotic condition of streams. A BioRecon method
is used at most of the sites; this is a more rapid sampling method and generates either a family-level or
genus-level assessment index. The semiquantitative method is more thorough and generates a
Tennessee Macroinvertebrate Index (TMI). Both methods are used at ecoregion refinement and
reference sites, while either method can be used at sites in other programs (depending on previous
scores received). For this Assessment, the family-level BioRecon Score was used because it was sampled
in more areas of the state and had a higher number sample count over time. From 1996 to 2014, the
TDEC database contains 5,369 BioRecon Scores, spread among the Level III ecoregions.
Fish IBI: The TVA fish IBI data used for this Assessment were sampled between 1996 and 2013. The most
recent sample date was used resulting in 789 records. TVA data were only collected in the Tennessee
and Cumberland River basins (Figure 1). TWRA collects the same fish IBI data in eastern Tennessee
(Ridge and Valley and Blue Ridge Mountains ecoregions). Because the same methods were used to
sample the fish and calculate the IBI, these datasets were combined to increase the sample size for this
analysis. Because of a low sample size in other ecoregions, fish IBI modeling was only performed in the
Ridge and Valley and Blue Ridge ecoregions of the state.
Tennessee Integrated Assessment of Watershed Health
32
-------
2.5 Watershed Vulnerability Metrics
Watershed vulnerability is defined as the potential for future degradation of watershed processes and
aquatic ecosystem health. Vulnerability within a watershed can be viewed as having two components:
(1) the possible threat, which is the process or event that causes a negative impact, and (2) resilience,
which is the sensitivity and adaptive capacity of the feature or process being impacted. The specific
threats that we examined for this Assessment were Land Use Vulnerability, Water Use Vulnerability, and
Climate Change Vulnerability (Figure 8). We evaluated the vulnerability of individual catchments to each
of these stressors. This section describes the watershed vulnerability metrics, the reasoning for their
selection, data sources, and methods applied to calculate metric values.
Figure 8. Watershed vulnerability metrics used for the Assessment.
Land Use
Projected Change
in Impervious
Cover
Potential for
Energy
Development
Water Use
Projected Change
in Water
Consumption
Projected Change
in Water
Withdrawals
Climate
Change
Increase in Days
Maximum
Temperature
>95°F
Increase in Days
without
Precipitation
Metrics of watershed vulnerability were rank normalized for reporting the metric, sub-index, and final
index calculations. Rank normalization transforms one or more variables to a uniform distribution and
scale, typically from 0 to 100; this common scale allows for comparisons between variables that may
exhibit different units and scales. Rank normalization is also insensitive to outlier or extreme values,
which can overly compress a normalized distribution when other normalization methodologies are
applied (Mitchell, 2012). This method is described in Appendix E.
2.5.1 Land Use Vulnerability Metrics
Natural land cover is important to protecting healthy watershed functions. However, the population in
Tennessee has been and is projected to continue to grow. This growth is located in urban areas with a
decreasing trend in agricultural and forested lands. Urban growth is associated with an increase in
impervious area. The resulting loss of natural land cover will increase the vulnerability of watersheds to
degradation. In addition to growing populations, energy development for economic growth is also a
Tennessee Integrated Assessment of Watershed Health
33
-------
potential threat to the landscape. The first step to assessing vulnerability due to land use change was to
determine where these changes were likely to occur.
Projected Impervious Cover Change: To determine vulnerability to increased urbanization, the
Assessment used percent impervious area projections produced by TNC (unpublished data). TNC used
population growth projections developed by the Tennessee Advisory Commission on Intergovernmental
Relations and the University of Tennessee Center for Business and Economic Research at 5-year intervals
through 2040. These population densities were converted to estimates of percent total impervious area
using the methods developed by the Greater Vancouver Sewerage and Drainage District and adopted by
EPA. The projected change in impervious area between 2010 and 2040 was calculated and those data
were used to determine this vulnerability metric. Additional information is available in the TN SWAP's
2012 Data and Methods Update. The percentage change in impervious area was determined for each
catchment and the results were rank-normalized with the highest increase in impervious area having the
greatest vulnerability.
Potential for Energy Development: Natural lands in Tennessee contain valuable energy resources. TNC,
in cooperation with the Appalachian Landscape Conservation Cooperative (LCC), developed a spatially
explicit model predicting the probability of coal mining, shale gas, and wind development (Dunscomb et
al., 2014). This data layer was intersected with the data layers representing natural lands (forests and
wetlands) to determine which natural lands could be vulnerable to development for energy resources.
The probability for energy development was categorized at four levels: highest risk (> 75%), some risk
(50%-75%), low risk (< 50%), and no risk (0%). Each NHD catchment was assigned a risk category based
on these data. Note that the data were not available for the Mississippi Alluvial Plains, Mississippi Valley
Loess Plain, and Southeastern Plains ecoregions.
2.5.2 Water Use Vulnerability Metrics
Humans can greatly affect a watershed's natural hydrologic regime by altering the stream network and
underlying aquifers in the form of surface and groundwater withdrawals. These alterations have
corresponding effects on the health of aquatic ecosystems. Future water demands will vary based on
population growth, changes in the design and operation of the thermoelectric power industry, and
expansion of agriculture, industry and mining. Vulnerability due to changing demands in water
withdrawals and consumptive water use from these withdrawals between 2010 and 2040 is captured at
the county level for this Water Use Vulnerability Metric.
The following steps were applied to calculate projected water use change for each county and are based
on methods used in a study for TV A by Bohac and Bowen (2012) when developing projected water use
for the Tennessee River watershed:
• County-level water withdrawal data by use sector for 2010 were obtained from the
USGS for the state of Tennessee. The withdrawal data for the state were compiled using
data from Memphis Light, Gas and Water, TDEC Division of Water Resources, TV A, and
the U.S. Army Corps of Engineers (USACE) (Maupin et al., 2014).
Tennessee Integrated Assessment of Watershed Health
34
-------
• Total water withdrawals for 2010 were determined as the summation of total
freshwater withdrawals (surface + groundwater) for the sectors of public supply,
irrigation-crop, irrigation-golf, livestock, aquaculture, industry, and thermoelectric.
• Consumptive water use for 2010 was calculated by sector by applying the net water
demand factor calculated by Bohac and Bowen (2012): thermoelectric 0.5%, industrial
6.5%, public supply 42.8% (also applied to irrigation-golf), and irrigation 100% (also
applied to livestock and aquaculture).
• Total water withdrawals for 2040 were determined by sector using projection factors
based on county-level population growth and irrigated acreage changes and generalized
change factors calculated by Bohac and Bowen (2012).
- Population projections for 2040 were obtained from the Tennessee State Data Center at the
University of Tennessee (http://tndata.utk.edu/sdcdemographics.htm). The county-level
population growth factor was calculated as the ratio of 2040 population to the 2010
population. This factor was applied to the water use sectors of public supply, industrial, and
irrigation-golf.
- Irrigated acreage by county was obtained for 2007 and 2012 from the 2012 Census of
Agriculture conducted by the U.S. Department of Agriculture, National Agricultural Statistics
Service. The ratio of change from 2007 to 2012 was applied to project changes into the
future for irrigation-crop, livestock, and aquaculture.
- Thermoelectric withdrawals in 2040 were calculated using the generalized factor of 31%
decrease estimated by Bohac and Bowen (2012).
• Consumptive water use for 2040 was calculated by sector by applying the net water
demand factor calculated by Bohac and Bowen (2012) for 2035 to the 2040 withdrawal
estimates calculated using the above steps: thermoelectric 2%, industrial 7.3%, public
supply 44.1% (also applied to irrigation-golf), and irrigation 100% (also applied to
livestock and aquaculture).
• Change in water withdrawals and water consumption by county were calculated as the
difference between the 2040 and 2010 estimates divided by the 2010 estimates. To
present as the Water Vulnerability Metric, these changes were each rank normalized.
2.5.3 Climate Change Vulnerability Metrics
Changes in climate affect aquatic ecosystem health through multiple avenues including hydrologic, land
form, and biologic alterations. Climate changes can take the form of different magnitudes, intensities, or
frequencies of precipitation and temperature events. It is possible for overall average conditions, as
measured by different climate metrics, to remain the same yet also have a completely different climate
of more extreme values (e.g., in storm intensity or frequency, in temperature extremes, or in geospatial
differences). The impact of climate-driven changes to aquatic ecosystem health depends on these
various aspects of the climate experienced across the state.
Tennessee Integrated Assessment of Watershed Health
35
-------
This Assessment uses downscaled global climate model data of projected changes in temperature and
precipitation to evaluate climate change across the state. To capture the different aspects of the climate
regime related to drought and heat intensity, which are assumed to have the largest impact on the
aquatic ecosystem health, two climate change vulnerability metrics were used:
• maximum number of consecutive days with zero to minimal (< 0.1 inch) rainfall across a
30-year period; and
• average annual number of days with maximum temperatures greater than 95°F across
the 30-year period.
Both of these metrics were assessed as a change between the historic 30-year period (1980-2010) and
the projected future 30-year period (2010-2040).
Data to calculate these metrics were obtained from the "Downscaled CMIP3 and CMIP5 Climate and
Hydrology Projections" (http://gdo-dcp.ucllnl.org/downscaled cmip projections/). The most extreme
emissions scenario (RCP 8.5) from the Coupled Model Intercomparison Project 5 (CMIP5) was chosen to
demonstrate a more extreme estimate of the potential climate changes (Taylor et al., 2012) in an
attempt to better highlight the more vulnerable areas. Because the climate vulnerability metrics chosen
relied on statistically downscaled, bias-corrected daily data (details of which can be found at http://gdo-
dcp.ucllnl.org/downscaled cmip proiections/techmemo/downscaled climate.pdf). a single global
circulation model was used rather than an ensemble forecast or an average value across multiple
models. The Coupled Physical Model GFDL CM3 run by the National Oceanic and Atmospheric
Administration (NOAA) Geophysical Fluid Dynamics Laboratory provided daily data for the selected time
periods at a 1/8 degree spatial resolution (~12 km by 12 km) across the state (Bureau of Reclamation,
2013). While the Climate Vulnerability Sub-index is provided at this resolution, values by catchment
were also determined for use in calculating the Watershed Vulnerability Index. Although catchments
within the immediate vicinity of one another received the same Climate Vulnerability Sub-index value,
the variability gradient across the state was large enough to provide useful information in the
Watershed Vulnerability Index.
Tennessee Integrated Assessment of Watershed Health
36
-------
3. Results and Discussion
This section presents the analytical results and maps illustrating scores for the Watershed Health and
Vulnerability Indices and the sub-indices for Landscape Condition, Hydrologic Condition, Geomorphic
Condition, Water Quality, Habitat Condition, Biological Condition, Land Use Vulnerability, Water Use
Vulnerability, and Climate Change Vulnerability (full-page maps of all sub-indices and metrics are
provided in Appendix A).
3.1 Watershed Health Index
Watershed Health Index scores are mapped in Figure 9. The highest scoring areas are in the Blue Ridge
and Appalachian Mountains in eastern Tennessee and scattered throughout the Interior Plateau in the
central part of the state. These areas are influenced by stable geomorphology, low deviation from
natural streamflow, and relatively high water quality and habitat conditions able to support diverse
biological communities. The lowest scoring areas are in the Ridge and Valley in eastern Tennessee and
the Mississippi Valley Loess Plains in western Tennessee. Land use in these regions is dominated by
agricultural and urban use, which alters the natural land cover and hydrology. These changes negatively
impact water quality and habitat leading to less diverse biological communities.
As described in Section 2, these scores were based on a collection of metrics that describe catchment land
cover and the physical, chemical, and biological attributes of stream ecosystems. Scores were quantified
from measured values of watershed health metrics (e.g., percent natural land cover, dam storage ratio)
and from statistical models of stream conditions (e.g., stream total phosphorus concentration). Modeled
metric values were based on a set of predictors that characterize both natural and anthropogenic
watershed features across multiple scales. Watershed Health Index scores therefore reflect ecological
condition gradients shaped by 1) natural variation in soils, topography, geology, hydrology, and similar
factors, 2) anthropogenic stressors that have influenced measured metric values, and 3) incremental and
cumulative scale anthropogenic stressors determined to be relevant to watershed health through
regression modeling. High scoring areas possess natural watershed characteristics that are shared by
healthy aquatic ecosystems and lack anthropogenic features associated with degraded ecosystem health.
3.2 Watershed Vulnerability Index
Watershed Vulnerability Index scores are mapped in Figure 10. Scores are highest in the southern Blue
Ridge Mountains in eastern Tennessee, northwestern Interior Plateau region of central Tennessee, and
Mississippi Valley Loess Plains in western Tennessee. Lowest vulnerability scores are in the Appalachian
Mountains and eastern portion of the Interior Plateau and along the northeastern Tennessee border.
Watershed Vulnerability Index scores present an approximation of the potential for future degradation
of aquatic ecosystem health. They depict projected exposure to climate, land use, and water use
change, but do not explicitly quantify how projected exposure translates to changes in the physical,
chemical, and biological makeup of a water body. The index is intended to be used in conjunction with
Watershed Health Index scores to identify areas that are currently healthy but vulnerable and most in
need of protection.
Tennessee Integrated Assessment of Watershed Health
37
-------
Figure 9. Watershed Health Index and sub-index scores for Tennessee
Tennessee Integrated Assessment of Watershed Health
N
A
Biologic Condition Sub-Index
38
-------
Figure 10. Watershed Vulnerability Index and sub-index scores for Tennessee.
Tennessee Integrated Assessment of Watershed Health
39
-------
4. Assumptions and Limitations
Assumptions were made throughout the development of this Assessment that may impose limitations
on using the results for certain watershed protection planning efforts. These assumptions should be
recognized by users of the Assessment output and are described below.
4.1 Spatial Framework
• The NHDPIus stream network is a medium-resolution (1:100,000) representation of
water body locations in Tennessee. Although the accuracy of the NHDPIus stream
network and catchment delineations were not verified as part of this project, they were
determined to be sufficient for regional screening of watershed protection priorities.
• Metric, sub-index, and index scores describe overall or average conditions within a given
NHDPIus catchment. Assessment results do not supply information at a resolution finer
than the catchment scale (approximately 1 square mile).
4.2 Watershed Health Metrics and Sub-Indices
• Watershed health metrics were selected on the basis of data availability, data quality,
spatial and temporal coverage, and expert judgment of relevance to watershed health.
Index scores do not account for aspects of watershed health beyond those represented
by selected metrics and the data from which they were derived.
• For statistical modeling, the Assessment assumed that the number and distribution of
samples was adequate for creating valid models predicting condition by ecoregion and
stream type (specifically that samples collected and metrics used in smaller streams
could be applied to larger streams where no field data may have been collected).
• Correlation among metrics was not factored into the metric selection process.
Correlation can suggest that one metric supplies "redundant" information that is
already provided by another metric, thus resulting in index scores weighted towards the
correlated metric results.
4.2.1 Landscape Condition Sub-Index
• The 2011 NLCD used in this Assessment was assumed to represent current landscape
conditions. In addition, the NLCD has a spatial resolution of 30 meters or 0.25 acres;
therefore, features or land use changes smaller than the minimum mapping unit were
not captured.
• The categorization of NLCD classifications as natural, semi-natural, and non-natural
was based on the descriptions of the classifications and agreed upon by the Technical
Team. For example, shrub/scrub is defined as areas dominated by trees generally less
than 16 feet (5 meters) tall and shrub canopy greater than 20% of the total vegetation
canopy. This class includes young trees in the early successional state or trees stunted
from environmental conditions. It is not until the tree height is greater than 16 feet
(5 meters) that the area would be classified as a forest. Because of the transitional state,
Tennessee Integrated Assessment of Watershed Health
40
-------
the shrub/scrub and grassland/herbaceous classes were considered semi-natural lands,
whereas forest lands were considered natural lands. Semi-natural lands represent a
gradient in departure from natural conditions, but do not degrade watershed health as
much as non-natural lands. Therefore, the semi-natural lands were weighted less than
natural lands but still included in the Landscape Condition metrics. This means that
catchments with semi-natural lands were ranked higher than catchments with a higher
percentage of non-natural lands.
• Forest lands identified as lands managed, owned, or operated for timber production
were classified as semi-natural lands for this Assessment. These managed forests differ
from natural forests in that they consist of a monoculture of often the same age and are
managed to maximize timber production through such activities as altered hydrology or
application of herbicides. When timber harvesting occurs, aquatic ecosystems are
stressed by increases in 1) water temperature range, 2) turbidity and sedimentation,
3) dissolved nutrients, 4) allochthonous organic detritus, and 5) streamflow (Lynch et al.,
1980; Swank et al., 1989).
4.2.2 Geomorphic Condition Sub-Index
• Geomorphology attempts to describe and quantify a variety of forces and processes
that form and shape dynamic river systems. The complexity of these channel-forming
processes, and the fact that even when undisturbed, rivers move across their
landscapes and are constantly being reshaped to some degree by the water flowing
through, makes it challenging to create a state-level tool for predicting geomorphic
stability. Typical field geomorphic measurements were not available at the catchment
scale and are very site-specific. Unlike water quality and biological monitoring,
geomorphic field assessment and monitoring are not performed frequently enough or at
a broad enough scale to provide field-based datasets that can be used to develop
statistical models to predict channel stability within unsurveyed catchments.
• Since a statistical model could not be developed, geospatial data were used to predict
potential Geomorphic Condition within a catchment. Some of the variables that
influence channel formation, such as bed roughness or the degree of a channel's
connection to its floodplain, cannot be determined from or be substituted with the
available hydrology, geologic, and landscape data. We focused on the factors that
determine channel stability including streamflow and land cover types associated with
runoff control and attenuation. The geospatial data used do not encompass all
components of geomorphology and unique local conditions may drastically alter the
character of individual catchments; therefore, the Geomorphic Condition Sub-index
should be considered a coarse estimator of likelihood of stream channel alteration.
4.2.3 Hydrologic Condition Sub-Index
• Hydrologic condition was assessed in three distinct pieces: eastern unregulated waters,
western unregulated waters, and statewide regulated waters. For the hydrologic
condition, reference sub-basins determined by high percentages of forest area
represent areas with least-disturbed streamflow. The equations created for the
Tennessee and Cumberland River basins (eastern Tennessee) were used for the small
Tennessee Integrated Assessment of Watershed Health
-------
portions of the Coosa basin (Alabama River) in the southeast portion of the state. The
equations for eastern Tennessee were based on a variety of physical factors
characterizing land use, slope, soils, subsurface, climate, and ecoregion. The equations
for western Tennessee relied on three factors: drainage area, geologic factors, and soil
factors, which were also used in the east (Appendix C).
4.2.4 Habitat Condition Sub-Index
• For Habitat Condition, the Assessment assumed that the RBP score was representative
of the habitat of the catchment. While the RBP integrates many habitat metrics into one
score, there are many components that are not measured. Habitat for individual species
or guilds cannot be addressed due to data limitations, thus the focus was on generalized
aquatic habitat condition.
4.2.5 Biological Condition Sub-Index
• Information was not available on all biological components of the ecosystem. For
example, consistent fish IBI data are not available for the entire state. Both TVA and
TWRA collect fish data, but IBI data was not reported for all sites. Therefore, there was
only enough data for the Ridge and Valley and Blue Ridge Mountains ecoregions to be
used in this Assessment. Other aquatic assemblages (mussels, periphyton) can be used
to assess watershed health, but adequate data on these assemblages were not
available. Samples were also limited in number and distribution by geography and
gradient of disturbance.
4.3 Watershed Vulnerability Metrics and Indices
• Metrics of watershed vulnerability were selected on the basis of data availability, data
quality, and expert judgment of relevance to watershed vulnerability. Index scores did
not account for aspects of watershed vulnerability beyond those represented by
selected metrics.
• Values of the projected impervious cover change metric reflect estimated changes in
impervious cover due to urban expansion only. Land use changes resulting from
agricultural expansion were not accounted for in the Assessment due to a lack of data.
• For water use vulnerability, individual power plants were not assessed for specific
planned changes in the future (i.e., conversion of a coal-fired steam plant to a combined
turbine plant); instead, a blanket rate of change to withdrawals and consumption was
used on all 2010 power plant water use data to predict future conditions. Additionally,
water use from interbasin transfers, irrigation, and aquaculture were not considered.
The rate of increase/decrease in irrigated agriculture by county from 2007 to 2012
remains constant from 2010 to 2040.
• Climate vulnerability was assessed through the results of a single global circulation model
scenario (created by NOAA) intended to represent a worst-case scenario.
Tennessee Integrated Assessment of Watershed Health
-------
5. Potential Applications of Assessment Results
This Assessment integrated many datasets to characterize watershed health and vulnerability across
Tennessee. The results are intended to support screening-level assessments of protection priorities and
are not intended to be used to determine the absolute condition of aquatic ecosystems. Results can also
serve as a baseline for evaluating change in watershed health over time and to assess the effectiveness
of existing protection strategies. In addition to this static report, the results are available as geospatial
data layers, which enables users to analyze the results at different spatial scales. The results presented
in this report are at the finest spatial resolution of individual catchments (median size of 0.6 square
miles), but the results could be aggregated up to HUC12 or HUC10 scale or within ecoregions. The
following is a summary of potential application of the Assessment results proposed by the THWI
Technical Committee.
Watershed Planning. This Assessment complements Tennessee's existing watershed-based approach to
resource management. Watershed planning occurs at a state, local, and regional level. The results from
this Assessment can provide a common framework to identify future watershed protection and
restoration goals. Results can also be used in conjunction with field observations to help determine
appropriate management actions in a watershed as part of the planning process.
Improved Monitoring and Assessment. Assessment results can inform aquatic ecosystem monitoring
programs that aim to collect data across a broad range of watershed conditions. Results can be used to
evaluate the range of watershed conditions currently monitored and to screen priority watersheds for
expanded monitoring. Results can also guide the selection of reference watersheds for developing
biological condition gradients and tracking changes in reference watersheds over time to validate the
effectiveness of watershed protection actions.
Outreach and Communication. Maps and other projects derived from the Assessment can communicate
information on the importance of watershed protection with nontechnical audiences and the public, as
well as gaining attention from national and regional decision makers. This information can support the
efforts of existing watershed protection organizations and identify where new organizations are needed.
Improved Decision Making. The identification of healthy intact watersheds could inform a variety of
decision-making processes including compensatory mitigation, land acquisition, and mine reclamation
projects. The results can foster cooperation across agencies and with other partners to protect priority
watersheds.
Economic Assessment. The Assessment results can be used as an input basis for conducting a
cost/benefit analysis focused on communicating the economic importance of protecting the most
ecologically healthy areas of Tennessee. The results are particularly useful for weighing the impact of
land use and water use decisions on health of aquatic ecosystems across the state.
Tennessee Integrated Assessment of Watershed Health
43
-------
Filling Data Gaps. The Assessment reflects a comprehensive inventory of available data for
characterizing watershed health and vulnerability. This process identified data gaps in current
monitoring programs and areas where different agencies could collaborate to improve comparability
methods. For example, while fish data are available across the state, there is not a standard method for
calculating I Bis. Therefore, Fish IBI scores were used as a metric only in the Ridge and Valley and Blue
Ridge Mountain ecoregions. Developing this Assessment has revealed opportunities like these for
improving incomplete datasets and creating others that can strengthen the Assessment and its
application.
Tennessee Integrated Assessment of Watershed Health
44
-------
6. References
Barnett, A. 2015. The Nature Conservancy. 100 Peachtree St. Atlanta, GA. E-mail Communication with
RTI. April 2015.
Bohac, C.E., and A.K. Bowen. 2012. Water Use in the Tennessee Valley for 2010 and Projected Use in
2035. Tennessee Valley Authority. 89 pp.
Bureau of Reclamation. 2013. Downscaled CMIP3 and CMIP5 Climate and Hydrology Projections: Release
of Downscaled CMIP5 Climate Projections, Comparison with preceding Information, and
Summary of User Needs. 47 pp. Denver, CO: U.S. Department of the Interior, Bureau of
Reclamation, Technical Services Center.
Castro, J., F. Reckendorf, and U.S. Natural Resources Conservation Service. 1995. RCA III, Effects of
Sediment on the Aquatic Environment: Potential NRCS Actions to Improve Aquatic Habitat. U.S.
Dunscomb J.K., J.S. Evans, J.M. Strager, M.P. Strager, and J.M. Kiesecker. 2014. Assessing Future Energy
Development across the Appalachian Landscape Conservation Cooperative. Charlottesville (VA):
The Nature Conservancy. 48 pp with appendices. Appalachian Landscape Conservation
Cooperative Grant #2012-02.
Griffith, G.E., J.M. Omernik, and S.H. Azevedo. 1997. Ecoregions of Tennessee. Corvallis, OR: U.S.
Environmental Protection Agency, EPA/600R-97/022, 51 p.
Jones, D.S., D.G. Kowalski, and R.D. Shaw. 1996. Calculating Reviewed Universal Soil Loss Equation
(RUSLE) Estimates on Department of Defense Lands: A Review ofRUSLE Factors and U.S. Army
Land Condition-Trend Analysis (LCTA) Data Gaps. Fort Collins, CO: Center for Ecological
Management of Military Lands. Department of Forest Science, Colorado State University.
Knight R.R., W.S. Gain, and W.J. Wolfe. 2012. Modelling ecological flow regime: an example from the
Tennessee and Cumberland River basins. Ecohydrology 5: 613-627.
Knight, R.R., J.C. Murphy, W.J. Wolfe, C.F. Saylor, and A.K. Wales. 2014. Ecological limit functions relating
fish community response to hydrologic departures of the ecological flow regime in the
Tennessee River basin, United States. Ecohydrology 7(5): 1292-1280.
Law, G.S., G.D. Tasker, and D.E. Ladd. 2009. Streamflow-Characteristic Estimation Methods for
Unregulated Streams of Tennessee. U.S. Geological Survey Scientific Investigations Report 2009-
5159, 212 p.
Leopold, L.B., M.G. Wolman, and J.P. Miller. 1964. Fluvial Processes in Geomorphology. W.H. Freeman
and Company. San Francisco, CA. 511 p.
Lynch, J.A., E.S. Corbett, and W.E. Sopper. 1980. Evaluation of management practices on the biological
and chemical characteristics of streamflow from forested watersheds. Inst, for Res. on Land and
Water Resources. PA St. U., University Park, PA.
Tennessee Integrated Assessment of Watershed Health
45
-------
Maupin, M.A., J.F Kenny, S.S. Hutson, J.K. Lovelace, N.L. Barber, and K.S. Linsey. 2014. Estimated Use of
Water in the United States in 2010. U.S. Geological Survey Circular 1405, 56 p.,
http://dx.doi.org/10.3133/cirl405.
Mitchell, H.B. 2012. Data Fusion: Concepts and Ideas. Springer, 2nd edition. 346 pages.
Nicholson, S.W., C.L. Dicken, J.D. Horton, K.A. Labay, M.P. Foose, and J.A.L. Mueller. 2005. Preliminary
integrated geologic map databases for the United States: Kentucky', Ohio, Tennessee, and West
Virginia. Version 1.1. U.S. Geological Survey Open-File Report 2005-1324, digital dataset.
Omernik, J., and G. Griffith. 2009. Ecoregions of Tennessee (EPA). Available at
http://www.eoearth.org/view/article/152206
Poff, N.L, J.D. Allan, M.B. Bain, J.R. Karr, K.L. Prestegaard, B.D. Richter, R.E. Sparks, and J.C. Stromberg.
1997. The natural flow regime: A paradigm for river conservation and restoration. Bioscience
47(11): 769-784.
Rosgen, D. 2006. Watershed Assessment of River Stability and Sediment Supply. Wildland Hydrology,
Fort Collins, CO.
Swank, W.T., L.F. DeBano, and D. Nelson. 1989. Effects of timber management practices on soil and
water. Pages 79-106 in R. Burns (Tech. comp.), The Scientific Basis for Silvicultural and
Management Decisions in the National Forest System. GTR-WO-55. Washington, DC. USDA
Forest Service.
Taylor, K.E., R.J. Stouffer, and G.A. Meehl. 2012. An Overview of CMIP5 and the experiment design.
Bulletin of the American Meteorological Society, 93: 485-498. doi:10.1175/BAMS-D-ll-00094.1
Tennessee Department of Environment and Conservation (TDEC). 2011. Quality System Standard
Operating Procedure for Macroinvertebrate Stream Surveys. State of Tennessee, Department of
Environment and Conservation, Division of Water Pollution Control.
Tennessee Healthy Watershed Initiative (THWI). 2015. Watershed Stewardship. Tennessee Department
of Environment and Conservation. Available at
https://www.tn.gov/environment/article/tennessee-healthy-watershed-initiative
The Nature Conservancy (TNC). 2012. Database Development and Spatial Analyses in Support of
Tennessee's State Wildlife Action Plan, 2012 Data and Methods Update. Available at
http://teaming.com/sites/default/files/TN-
SWAP%20Data%20and%20Methods%20Update%20Report%202012.pdf.
U.S. Environmental Protection Agency (USEPA). 2012. Identifying and Protecting Healthy Watersheds:
Concepts, Assessments, and Management Approaches. EPA 841-B-11-002. Washington, DC: U.S.
Environmental Protection Agency.
U.S. Environmental Protection Agency (USEPA). 2014. Watershed Index Online (WSIO). Available at
http://gispub.epa.gov/wsio/.
Tennessee Integrated Assessment of Watershed Health
46
-------
Appendix A: Map Atlas
This appendix contains full page maps for all watershed health and vulnerability metrics, sub-indices,
and indices for all National Hydrography Dataset Plus (NHDPIus) catchments in Tennessee. The following
guidelines were used for map development:
• Maps display rank-normalized metric, sub-index, or index scores.
• Maps were created using 10 equal-interval color classes. Because scores are rank-
normalized, these classes generally correspond to deciles.
• Maps display metrics in their directionally aligned scores used for sub-index and index
calculations rather than original directionality (see Table E-l). For example, catchments
with low total nitrogen concentrations have a high affinity for watershed health and
therefore are scored high for the total nitrogen metric.
Tennessee Integrated Assessment of Watershed Health
A-l
-------
Figure A-l. Watershed Health Index.
Figure A-2. Landscape Condition Sub-Index.
Tennessee Integrated Assessment of Watershed Health
A-2
-------
Figure A-3. Landscape Condition Metric: Percent natural land cover.
Figure A-4. Landscape Condition Metric: Percent natural land in hydrologically active zone.
Tennessee Integrated Assessment of Watershed Health
A-3
-------
Figure A-5. Geomorphic Condition Sub-Index.
Figure A-S. Geomorphic Condition Metric: Erosive forces.
Tennessee Integrated Assessment of Watershed Health
A-4
-------
Figure A-7. Geomorphic Condition Metric: Resistive forces.
Figure A-S. Hydrologic Condition Sub-Index.
Tennessee Integrated Assessment of Watershed Health
A-5
-------
Figure A-9. Hydrologic Condition Metric: Streamflow characteristics deviation from reference condition (unregulated streams).
Figure A-10. Hydrologic Condition Metric: Dam storage ratio (regulated streams).
Tennessee Integrated Assessment of Watershed Health
A-6
-------
Figure A-ll. Water Quality Sub-index.
Figure A-12. Water Quality Metric: Stream total nitrogen condition. Note: High condition values indicate relatively lower total nitrogen
concentrations.
Tennessee Integrated Assessment of Watershed Health
Total Nitrogen
Low
-------
Figure A-13. Water Quality Metric: Stream total phosphorus condition. Note: High condition values indicate relatively lower total
phosphorus concentrations.
Figure A-14. Water Quality Metric: Stream specific conductance condition. Note: High condition values indicate relatively lower specific
conductance values.
Tennessee Integrated Assessment of Watershed Health
A-8
-------
Figure A-15. Habitat Condition Sub-Index.
Figure A-16. Rapid Bioassessment Protocol (RBP) scores.
Tennessee Integrated Assessment of Watershed Health
A-9
-------
Figure A-17. Habitat Suitability scores.
'T V*
/
<" i"™, f—~
i:
- K i*'
> \ A '" j,
: i,V/ , 1 >
ry-
"i ¦ • v »• „.=>>" t *
# , Vi5*' ^ f" \ ** >W**p ' A
*¦ > 'v»£5L ' " "• v *
* a* > V * ¦* V N
«*v /
-^&L d i
Habitat Suitable Score
High^^H Very High I
]Miles
Figure A-18. Biological Condition Sub-Index.
Tennessee Integrated Assessment of Watershed Health
A-10
-------
Figure A-19. Biological Condition Metric: Benthic macroinvertebrate Index of Biological Integrity (IB!) rating.
Figure A-20. Biological Condition Metric: Fish Index of Biological Integrity (IBI) rating.
Tennessee Integrated Assessment of Watershed Health
A-ll
Fish IBI
Low
-------
Figure A-21. Watershed Vulnerability Index.
Figure A-22. Land Use Vulnerability Sub-Index.
Tennessee Integrated Assessment of Watershed Health
A-12
-------
Figure A-23. Land Use Vulnerability Metric: Projected change in impervious cover.
Figure A-24. Land Use Vulnerability Metric: Potential for energy development.
IV-
9
r
r
Potential for Energy Development
Low
75 100
I Miles
High
Tennessee Integrated Assessment of Watershed Health
A-13
-------
Figure A-25. Water Use Vulnerability Sub-Index.
N
A
0 25 50
75 100
¦ I Miles
Water Use Vulnerability Sub-Index
Low
High
Figure A-26. Water Use Vulnerability Metric: Projected change in water consumption.
Tennessee Integrated Assessment of Watershed Health
A-14
-------
Figure A-27. Water Use Vulnerability Metric: Projected change in water withdrawals.
[Miles
Projected Changes in Water Withdrawals
Figure A-28. Climate Change Vulnerability Sub-Index.
75 100
Miles
Climate Vulnerability Sub-Index
Tennessee Integrated Assessment of Watershed Health
A-15
-------
Figure A29. Climate Change Vulnerability Metric: Increase in days with maximum temperature greater than 95°F.
Increase in Average Annual Number of Days with
Temperature > 95 F
Low
High
Figure A-30. Climate Change Vulnerability Metric: Increase in consecutive days without precipitation (> 0.1 inch).
75 100
~lMilps
Increase in Average Drought Length
Low
No Increase
High
Tennessee Integrated Assessment of Watershed Health
A-16
-------
Appendix B: Geomorphic Condition
B.l Introduction
A combination of stream channel characteristics and landscape conditions derived primarily from the
2011 National Land Cover Database (NLCD 2011) was used to predict the Geomorphic Condition of
catchments within Tennessee. A simple comparative analysis was used to attempt to gauge the relative
potential for channel stability and resiliency among the catchments.
Separate metric analyses were performed to evaluate the erosive forces and the resistive capacity
within stream channels for each catchment. Then elements of the two were combined to determine
overall Geomorphic Condition.
B.2 Preparation of Data and Geomorphic Condition Sub-Index
Landscape factors, such as percent impervious cover and percent forest cover (for the cumulative land
area draining to the catchment), and percent natural land cover within the Hydrologically Active Zone
(HAZ), were calculated using NLCD 2011 datasets. Other variables, such as K-factor and depth to bedrock
were obtained from National Hydrography Dataset Plus (NHDPIus), Soil Survey Geographic Database
(SSURGO), and U.S. Geological Survey (USGS) datasets. The streamflow factors used for this analysis
were taken from the modeling run for the Hydrologic Condition Sub-Index.
A simple continuous scoring model (range: 0-100 points) was used for each factor included in the
analysis (Tables B-l). Each was scored so that higher point values indicated the factor would have a
positive effect on stream resilience and stability (e.g., a lower percentage of impervious cover is less
likely to alter the natural flow regime in a catchment and therefore would score more points than a
catchment with a higher percentage of impervious cover).
Table B-l. Scoring components of the Geomorphic Condition Sub-Index.
Factor
Minimum Value (Point
Score)
Maximum Value (Point Score)
Erosion Metric
% Impervious Cover (cumulative)
0% (100 points)
65.06% (0 points)
K-Factor
0.00006811 (100 points)
0.55 (0 points)
Eastern Tennessee - Mean Annual Runoff
0.535 (100 points)
4.542 (0 points)
Western Tennessee - Mean Summer Streamflow
0.00 (100 points)
0.85 (0 points)
Resistance Metric
% Forest Cover (cumulative)
0% (0 points)
100% (100 points)
Depth to Bedrock (cm)
0 (100 points)
153 (0 points)
% Natural Cover in HAZ (local)
0% (0 points)
100% (100 points)
1 Note that K-factor values of less than 0.02 were possible in this analysis due to this value being a calculated, length-
weighted average
Tennessee Integrated Assessment of Watershed Health
B-l
-------
The values for each factor were assigned points by rescaling the range of actual values for that factor
from 0 to 100. The three factor scores were then averaged to produce a metric score. If a value for a
particular factor was not available, this factor was dropped from the average. In this manner, each
catchment was given an erosion metric score and a resistance metric score. The scores for each
respective metric were then rank normalized. The rank normalized metric scores were then averaged
and that average was rank normalized to produce a Geomorphic Condition Sub-Index score.
B.3 Model Development
Experience with a previous large-scale geomorphic condition assessment performed under the Healthy
Watersheds Program helped guide the approach to and development of this model. Previous attempts
to classify catchments based on Rosgen stream type proved untenable with the datasets available at the
broad, statewide scale, so that approach was not attempted for this model.
Various discussions were had regarding complicating factors within the state, such as karst geology and
the inconsistent availability of flow metrics on a statewide basis. Geographic datasets for karst geology
were obtained from USGS; however, the data were too coarse for targeted application of karst data as
either an additional metric factor or as a sieve for eliminating catchments from evaluation where the
model might be rendered less effective due to the effects of the karst. Discussions with the Tennessee
Healthy Watershed Initiative also raised the counterpoint that perhaps the impacts of karst were
isolated enough as to be a minor concern for the statewide assessment.
B.4 Model Evaluation
Sampling and surveys performed for water quality and stream biota, although not always
comprehensive for every basin or watershed within a state the size of Tennessee, typically cover a great
number of locations and can be used to generalize reaches of stream (or even entire catchments)
beyond where they occur. Geomorphological surveys are not performed as often or at as many
locations, and the type of data collected are extremely site specific. This makes it difficult to apply data
collected in one area to others or develop broader statistical models based on the types of stream
channel measurements that are available. For this reason rather than attempting to use such a limited
dataset to make weak statistical correlations, a different approach was used.
Many of the processes and forces that contribute to channel stability are well known and understood.
Based on knowledge of these processes and the available data, best professional judgment was used to
select factors that would approximate the forces that lead to or resist the erosion that influences
channel morphology.
One of the strengths of this analysis is that it was based on data that were generally available for all
catchments within Tennessee. Although localized conditions may lead to conditions in individual
catchments that disagree with this Assessment, overall it is felt to be a good representation of the
relative potential channel stability among all of the catchments.
Tennessee Integrated Assessment of Watershed Health
B-2
-------
Appendix C: Hydrologic Condition
A stream's flow regime refers to its characteristic pattern of flow magnitude, timing, frequency,
duration, and rate of change (Poff et al., 1997). The flow regime plays a central role in shaping aquatic
ecosystems and the health of biological communities. Alteration of natural flow regimes (e.g., more
frequent floods) can reduce the quantity and quality of aquatic habitat, degrade aquatic life, and result
in the loss of ecosystem services. Therefore, to assess Hydrologic Condition, we use metrics related to
the flow regime in unregulated streams to determine which segments most closely resemble the natural
flow regime through reference watersheds and are therefore assumed to be healthy. In regulated
systems (i.e., streams below large dams), we use the ratio of the storage behind the dams to the
expected mean annual natural streamflow to determine which regulated segments have lower volumes
of storage compared to streamflow.
C.l Unregulated Streams
Independent variables used in the regression models for streamflow characteristics (SFCs) represent
four categories (i.e., climate, land use, physical landscape, and regional indicators) and are listed in
Table C-l. The independent variables either represent an average value or a percentage of land area
across the drainage area of each site.
To use these regression equations to determine the deviation of the flow regime from a natural state
within each catchment, the numeric value calculated for each utilized SFC was compared to a reference
range determined for each SFC. The deviation was assumed to be zero if the calculated SFC fell within
the reference range. For values outside the reference range, the absolute deviation was calculated as
the difference between the SFC value and the lower or the upper bound of the reference range
(depending on where the value falls). For each catchment, all absolute deviations for the used SFCs were
then added together to calculate a single deviation metric (Figure C-l).
Figure C-l. Calculation method for Hydrologic Condition using deviation from reference range for
streamflow characteristics (SFCs).
C
D
~a
Q)
T3
C
ro
SFC1
Total Deviation = D1 + D2
SFC2 SFC3
~
i
Reference
Range
Site Value
Deviation
Tennessee Integrated Assessment of Watershed Health
C-l
-------
In the east, the reference ranges vary by ecoregion (Table C-2). These reference ranges are based on the
interquartile range of values across all monitoring sites used in the analysis and are differentiated
between the Blue Ridge, Ridge and Valley, and Interior Plateau ecoregions. (Note that these calculations
are completed with normalized values.) Catchments that have sub-basins with forest land use in the
upper quartile of the range of values (Table C-2) were used in calculating the interquartile range.
Table C-l. Definitions for independent variables used in predictive equations (adapted from Knight
et al.; 2012).
Variable
E
V
Definition (Data Source)
Variation from USGS
Study
Climate
Monthly Mean
Precipitation
E
Average annual precipitation divided by 12 (PRISM,
2004)
Same
January Precipitation
Deviation
E
Mean January precipitation divided by monthly
precipitation mean (PRISM, 2004)
Same
Daily Temperature Range
E
Mean maximum daily temperature minus mean
minimum daily temperature (PRISM, 2004)
Same
August Temperature
Deviation
E
Mean August maximum temperature minus mean
annual temperature divided by mean annual
temperature (PRISM, 2004)
Same
Land Use
Percent Forest Cover
E
The total percentage of land cover in a catchment that is
considered to be forested (Jin et al., 2013)
2011 land use in place of
2001 land use dataset
Percent Agricultural
Cover
E
The total percentage of land cover in a catchment that is
considered to be agricultural (Jin et al., 2013)
2011 land use in place of
2001 land use dataset
Physical
Horton
E
Index of Hortonian overland (infiltration excess overland
flow) (dimensionless) (Wolock et al., 2003a and 2003b)
Same
Mean Elevation
E
Mean basin elevation derived from digital elevation
model (Gesch et al., 2002; Gesch, 2007)
Derived from more recent
digital elevation dataset
Soil Factor
E, W
Percentage of area underlain by soil with a permeability
of at least 5 cm/h (Greene and Wolfe, 1998)
Same
Rock Depth
E
Average depth of soil above bedrock (Wolock, 1997)
Same
Regional
Geologic Factor
E, W
Measure of the number of days that pass as discharge
recedes one complete log cycle of streamflow (days)
(Bingham, 1986)
Same
Blue Ridge
E
Percentage of the watershed that lies within the Blue
Ridge Level 3 ecoregion (Omernik, 1987)
Same
Interior Plateau
E
Percentage of the watershed that lies within the Interior
Plateau Level 3 ecoregion (Omernik, 1987)
Same
Interaction Terms
Soil Factor
E
Soil factor multiplied by monthly mean precipitation
N/A
Rock Depth
E
Rock depth multiplied by monthly mean precipitation
N/A
Geologic Factor
E
Geologic factor multiplied by monthly mean
precipitation
N/A
All variables represent average values for a basin with the exception of Blue Ridge, Interior Plateau, forest, and agriculture,
which are expressed as the percentage of total catchment area.
Tennessee Integrated Assessment of Watershed Health
C-2
-------
Table C-2. Interquartile ranges and forest land use threshold for ecoregion-specific hydrologic
reference profiles calculated for eastern Tennessee.
MA41
TL1
LDH16
Ecoregion
Forest %
25th
75th
25th
75th
25th
75th
Blue Ridge
97.1
0.1507
2.2775
-0.2672
0.5900
-0.4428
0.6654
Central Appalachians
90.7
-0.3240
0.7831
-0.6850
-0.4541
-0.7015
0.0140
Interior Plateau
73.3
-0.7080
-0.3763
-0.1712
0.5762
-1.9631
-1.4628
Ridge and Valley
66.4
-0.9566
-0.4065
-0.7996
-0.2818
-0.8873
-0.0640
Southeastern Plains
69.6
-0.5368
-0.1553
-0.1948
0.8101
-2.2207
-1.6972
Southwestern
Appalachians
80.1
-0.3421
0.5283
-0.9880
-0.4238
-1.1340
-0.7879
MA41 = mean annual runoff; TL1 = date of annual minimum flow; LDH16 = variability in high-pulse duration; Values are
normalized by mean and standard deviation
In the west, there is only one reference range for each SFC, which is again defined as the interquartile
range of reference catchments. For the west, reference catchments were those that had forest land use
in the upper quartile of the range of values (Table C-3).
Table C-3. Interquartile ranges and agriculture land use threshold for hydrologic reference profiles
calculated for western Tennessee.
7Q10
MS
qlO
Forest %
25th
75th
25th
75th
25th
75th
45.2
0.0362
0.1543
0.4146
0.6385
2.7424
3.0796
7Q10 = lowest consecutive 7-day average flow in a 10-year period; MS = mean-summer streamflow in June through August; qlO
= daily mean streamflow exceeded 10% of the time; Regression values are normalized by drainage area
C.2 Regulated Streams
Regulated streams were assessed by first identifying all large dams within the hydrologic units that cross
the state. Dams were restricted to those that stored at least 10,000 acre-feet of volume according to the
TN SWAP dataset, were not primarily for recreational use, and had a National Inventory of Dams (NID)-
reported normal storage above zero (Figure C-2). The final dams were compared with listings in Bohac
and Bowen (2012) and Robinson (2014) for the Tennessee and Cumberland River systems, respectively,
to ensure all large operated dams were selected using these criteria.
Tennessee Integrated Assessment of Watershed Health
C-3
-------
Figure C-2. Dams with greater than 10,000 acre-feet of storage within the hydrologic units crossing
Tennessee. Numeric labels relate to the ID field in Table C-4.
Legend
¦ Dams > 10,000 aere-fl Green ~ Upper Tennessee ( I Lower Tennessee
HUCSs I | Halchie-Obion | j Middle Tennessee-Elk | j French Broad-Holstofi
Coosa-Tallapoosa Lower Mississippi-Memphis [ ] Middle Tennessee-Hiwassee Upper Cumberland
Lower Cumberland
Information on each darn was gathered from TN SWAP, NID, and generalized searches on the owning
organizations (Table C-4). Attempts were made to qualify each dam based on the operating principles to
potentially provide a secondary metric in the ranking of regulated streams. Because the information was
only readily available for the larger management organizations (e.g., TVA) and a method of
standardization/categorization of the different guide curves and operating levels to calculate a
hydrologic metric was not objectively available given the screening level focus of this analysis, these
data are instead provided in tabular form and are not used in metric rankings.
C.3 References
Bingham, R.H. 1986. Regionalization of Low-Flow Characteristics of Tennessee Streams. Water-Resources
Investigations Report 85-4191, 63.2 plates.
Bohac, C.E. and A.K. Bowen. 2012. Water use in the Tennessee Valley for 2010 and projected use in
2035. Tennessee Valley Authority, River Operations and Renewables. 83 pp.
Gesch, D., M. Oimoen, S. Greenlee, C. Nelson, M. Steuck, and D. Tyler. 2002. The national elevation
dataset. Photogrammetric Engineering and Remote Sensing 68: 5-11.
Gesch, D.B. 2007. The national elevation dataset. In Digital Elevation Model Technologies and
Applications: The DEM Users Manual2nd edition, Maune, D. (ed). American Society for
Photogrammetry and Remote Sensing: Bethesda, MD; 99-118.
Tennessee Integrated Assessment of Watershed Health
C-4
-------
Greene, D.C., and W.J. Wolfe. 1998. Superfund GIS—Soil thickness, permeability, texture, and
classification in Tennessee. Available at http://catalog.data.gov/dataset.
Jin, S., L. Yang, P. Danielson, C. Homer, J. Fry, and G. Xian. 2013. A comprehensive change detection
method for updating the National Land Cover Database to circa 2011. Remote Sensing of
Environment 132:159-175.
Knight, R.R., W.S. Gain, and W.J. Wolfe. 2012. Modelling ecological flow regime: an example from the
Tennessee and Cumberland River basins. Ecohydrology 5: 613-627.
Omernik, J.M. 1987. Ecoregions of the conterminous United States. Annals of the Association of
American Geographers 77: 118-125.
Poff, N.L, J.D. Allan, M.B. Bain, J.R. Karr, K.L. Prestegaard, B.D. Richter, R.E. Sparks, and J.C. Stromberg.
1997. The natural flow regime: A paradigm for river conservation and restoration. Bioscience
47(11): 769-784.
PRISM Climate Group, Oregon State University, http://prism.oregonstate.edu. created 4 Feb 2004.
Robinson, J.A. 2014 (draft). Estimated use of water in the Cumberland River Watershed in 2010 and
projections to 2040: U.S. Geological Survey Scientific Investigations Report 14-xxxx, xx p.
Available online at http:pubs.usgs.gov/circ/14-xxxx.
Wolock, D.M. 1997. STATSGO soil characteristics for the conterminous United States. U.S. Geological
Survey Open-File Report 97-656. Digital dataset.
Wolock, D.M. 2003a. Saturation overland flow estimated by TOPMODEL for the conterminous United
States. U.S. Geological Survey Open-File Report 2003-264, raster digital data.
Wolock, D.M. 2003b. Infiltration-excess overland flow estimated by TOPMODEL for the conterminous
United States. U.S. Geological Survey Open-File Report 03-310, digital dataset.
Tennessee Integrated Assessment of Watershed Health
C-5
-------
Table C-4. Dams used to assess regulated stream reaches for the hydrologic basins within Tennessee.
NIDID
ID
Dam Name
Owner Name
Primary
Purpose
Hydrologic Operations Notes1
Water
Quality
Operations2
NID
Storage
(TN
SWAP)
Normal
Storage
Max
Storage
AL05901
1
Bear Creek
TVA
Flood Control
Guide curve
No
78580
9915
78580
AL05902
4
Cedar Creek
TVA
Flood Control
Guide curve, Low notch
No
173490
93201
173490
AL05903
2
Little Bear Creek
TVA
Flood Control
Guide curve, Low notch
No
74855
45609
74855
AL07701
3
Wheeler
TVA
Flood Control
Operating range
No
1358355
1049007
1358355
AL07702
6
Wilson
TVA
Flood Control
Operating range
No
674220
636543
674220
AL09301
5
Upper Bear Creek
TVA
Flood Control
Spillway crest, Low Notch,
Recreation release schedule
No
69810
37677
69810
AL09501
0
Guntersville
TVA
Flood Control
Operating range
No
1405947
1019262
1405947
GA00730
7
Upper Haig Mill Lake
Dam
City of Dalton
Flood Control
27728
483
27728
GA11101
9
Blue Ridge
TVA
Hydroelectric
Balancing guide, flood guide
with expected elevation range
Yes
228045
182436
228045
GA29101
10
Nottely
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
216147
162606
216147
KY00088
16
Wood Creek Lake Dam
Commonwealth
of Kentucky
Water Supply
29101
23270
29101
KY00275
33
Cannon Creek Dam
Commonwealth
of Kentucky
Water Supply
11300
11300
0
KY03001
14
Barkley Dam
USACE -
Nashville District
Hydroelectric
Guide curve
Yes
2082000
869000
2082000
KY03010
15
Wolf Creek
USACE -
Nashville District
Hydroelectric
Upper and lower guide curves,
minimum power pool
Yes
6089000
2142000
6089000
KY03046
11
Laurel Dam
USACE -
Nashville District
Flood Control
Power marketing curve
No
435600
185000
435600
KY03061
12
Martins Fork Dam
USACE -
Nashville District
Flood Control
Seasonal operating curve
No
21100
3700
21100
KY15701
13
Kentucky
TVA
Flood Control
Operating range, Summer
Flowage Easement
No
7535400
6127470
7535400
NC00181
24
Appalachia
TVA
Hydroelectric
Operating range
Yes
63456
55524
63456
NC00288
42
North Fork Reservoir
Dam
City of Asheville
Department of
Water Resources
Water Supply
21700
17600
21700
NC00298
21
Fontana
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
1552689
1370253
1552689
(continued)
Tennessee Integrated Assessment of Watershed Health
C-6
-------
Table C-4. Dams used to assess regulated stream reaches for the hydrologic basins within Tennessee, (continued)
NID
Water
Storage
Primary
Quality
(TN
Normal
Max
NIDID
ID
Dam Name
Owner Name
Purpose
Hydrologic Operations Notes1
Operations2
SWAP)
Storage
Storage
NC00318
19
Walters
Duke Energy
Carolinas, LLC
Hydroelectric
Recreation release schedule
17000
17000
17000
NC00336
26
Bear Creek
Duke Energy
Carolinas, LLC
Hydroelectric
Guide curve, min and max
curves, low inflow protocol
34600
34600
34600
NC00371
28/
30
Dicks Creek
Dam/Nantahala/Whit
e Oak Creek Dam
Duke Energy
Carolinas, LLC
Hydroelectric
Guide curve, min and max
curves, low inflow protocol
126000
126000
126000
NC00378
25/
29
Glenville Saddle
Dike/Thorpe
Duke Energy
Carolinas, LLC
Hydroelectric
Guide curve, min and max
curves, low inflow protocol
67100
65600
67100
NC00391
44
Chatuge
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
285552
226062
285552
NC00392
17
Santeetlah
ALCOA Power
Generating Inc.,
TAPOCO Division
Hydroelectric
Guide curve, release schedule
207000
160000
207000
NC00419
18
Hiwassee
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
471954
398583
471954
TN00904
70
Chilhowee
ALCOA Power
Generating Inc.,
TAPOCO Division
Hydroelectric
Run-of-river
49251
49251
49251
TN00906
62/
75
Calderwood
ALCOA Power
Generating Inc.,
TAPOCO Division
Hydroelectric
Run-of-river
43500
41100
43500
TN01302
66
Norris
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range,
recreation release schedule
Yes
3363168
2040507
3363168
TN01903
58
Watauga
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range,
recreation release schedule
Yes
751557
569121
751557
TN02101
68
Cheatham Dam
USACE -
Nashville District
Hydroelectric
No guides
Yes
104000
84200
104000
TN02702
57
Dale Hollow Dam
USACE -
Nashville District
Hydroelectric
Upper and lower guide curves,
minimum power pool
Yes
1706000
857000
1706000
TN03107
63
Normandy
TVA
Hydroelectric
Guide curve
No
126000
116997
126000
(continued)
Tennessee Integrated Assessment of Watershed Health
C-7
-------
Table C-4. Dams used to assess regulated stream reaches for the hydrologic basins within Tennessee, (continued)
NIDID
ID
Dam Name
Owner Name
Primary
Purpose
Hydrologic Operations Notes1
Water
Quality
Operations2
NID
Storage
(TN
SWAP)
Normal
Storage
Max
Storage
TN03504
59
Lake Tansi
Lake Tansi Village
P.O.A.
Other
13806
9000
13806
TN03701
67
J Percy Priest Dam
USACE -
Nashville District
Flood Control
Upper and lower guide curves
Yes
652000
202000
652000
TN03702
71
Old Hickory Dam
USACE -
Nashville District
Hydroelectric
Minimum power pool
Yes
545000
420000
545000
TN04102
55
Center Hill Dam
USACE -
Nashville District
Flood Control
Upper and lower guide curves,
minimum power pool
yes
2092000
1330000
2092000
TN05101
76
Elk River Dam
U.S. Air Force Air
Force Materiel
Command
Flood Control
101844
77915
101844
TN05102
78
Tims Ford
TVA
Flood Control
Guide curve, Flood guide,
Minimum Recreation level,
Recreation release schedule
Yes
608000
325400
608000
TN05903
74
Nolichucky
TVA
Other
No information
19525
1715
19525
TN06504
69
Chickamauga
TVA
Flood Control
Operating range
No
943908
622662
943908
TN07101
61
Pickwick Landing
TVA
Flood Control
Operating range
No
1546740
1118412
1546740
TN07305
38
John Sevier
TVA
Water Supply
No information
52650
7735
52650
TN07705
77
Beech
TVA
Flood Control
No information
28602
11105
28602
TN07706
32
Pine
TVA
Flood Control
No information
12260
5155
12260
TN07710
39
Pin Oak
TVA
Flood Control
No information
13815
8925
13815
TN08903
52
Cherokee
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
1699431
1421811
1699431
TN10501
64
Fort Loudoun
TVA
Flood Control
Operating range
No
475920
362889
475920
TN10502
37
Melton Hill
TVA
Hydroelectric
Operating range
No
150708
105099
150708
TN10506
65
Tellico
TVA
Flood Control
Operating range
No
513597
392634
513597
TN11502
50
Nickajack
TVA
Hydroelectric
Operating range, Bottom of
operating zone during high
flows
No
402549
246130
402549
TN11929
36
Solutia #15
Rlf Duck River,
LLC, etal.
-
32945
23614
32945
TN12102
47
Watts Bar
TVA
Flood Control
Operating range
No
1415862
1009347
1415862
TN13905
34
Ocoee No. 1
TVA
Flood Control
Guide curve, recreation release
schedule at Ocoee 2 and 3
No
79320
79320
48350
(continued)
Tennessee Integrated Assessment of Watershed Health
C-8
-------
Table C-4. Dams used to assess regulated stream reaches for the hydrologic basins within Tennessee, (continued)
NID
Water
Storage
Primary
Quality
(TN
Normal
Max
NIDID
ID
Dam Name
Owner Name
Purpose
Hydrologic Operations Notes1
Operations2
SWAP)
Storage
Storage
TN15501
72
Douglas
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
1626060
1223511
1626060
TN15901
51
Cordell Hull Dam
USACE -
Nashville District
Hydroelectric
Upper and lower guide curves
310900
258000
310900
TN16305
49
South Holston
TVA
Flood Control
Balancing guide, flood guide
with expected elevation range
Yes
890367
658356
890367
TN16306
48
Boone
TVA
Flood Control
Guide curve, flood guide
Yes
216147
180453
216147
TN16307
35
Fort Patrick Henry
TVA
Flood Control
Operating range
Yes
31728
25779
31728
TN17704
31
Great Falls
TVA
Hydroelectric
Guide curve
No
64800
39660
64800
Sources for operating notes:
http://www.lrn-wc.usace.armv.mil/hh/WM Info.htm
http://www.brookfieldrenewable.com/content/smokv mountain hydro/recreation and flow-39611.html
http://www.duke-enerev.com/lakes/nantahala/nantahala-lake-levels.asp
http://www.duke-energv.com/power-plants/hvdro/walters.asp
2Source for water quality operations:
http://www.tva.gov/environment/water/rri triblist.htmffnottelv
Bullard, personal communication via comments: 11 September 2015.
Tennessee Integrated Assessment of Watershed Health
C-9
-------
Appendix D: Water Quality, Habitat and Biological Condition Metric Modeling
D.l Introduction
Watershed health metrics characterizing water quality and habitat and biological condition were
quantified for each National Hydrography Dataset Plus (NHDPIus) catchment in Tennessee using
Boosted Regression Trees (BRTs). This statistical modeling approach was used to relate observed values
to landscape- and watershed-scale predictor variables. The fitted statistical models were then used to
predict water quality, habitat condition, and biological conditions in catchments without available
monitoring data.
D.2 General Modeling Approach
BRTs are a form of tree-based modeling and can be visualized as a series of nodes and branches that
represent partitions of the predictor data space into rectangular sections; each binary split partitions the
dataset into groups that are more rather than less similar in terms of the response variable (Cutler et al
2007). This partitioning is recursive; each additional split is added to previous splits until all observations
are partitioned or until some other stopping criteria is reached. For instance, in Figure D-l observations
with a 'Var3' value less than 8.5 are partitioned in a terminal node to the right, while observations with a
'Var3' value greater than or equal to 8.5 are directed down the left branch of the tree. Each split
represents a predictor variable threshold; a highly influential predictor may be split several times at
different nodes.
Figure D-l. Example decision tree output.
Var3 >*
/
/
< 8.5
Tennessee Integrated Assessment of Watershed Health
D-l
-------
BRT incorporates decision tree modeling with additional features to fit both continuous and categorical
response variables. BRT is a stagewise, ensemble modeling approach that utilizes both stochastic data
selection and boosting to improve predictive performance. A BRT model is a collection of hundreds or
thousands of individual trees that are combined to obtain the final fitted values, hence ensemble. The
BRT procedure is stagewise because each tree is fitted to the error term of the preceding tree and all
trees are retained in their original form. This emphasis on observations that are modeled poorly is a
form of boosting, which seeks to improve model performance by gradually reducing the overall error of
the assembled trees (Elith et al., 2008). Finally, many tree-based modeling approaches adopt a
stochastic, or random, element to the fitting process to improve overall performance. In this case, a 'bag
fraction' specifies the proportion of data that is drawn at each tree fitting step.
BRT models require user input on three settings: bag fraction, learning rate, and tree complexity. Default
values for these settings are provided; however, model performance can frequently be improved by
calibrating these arguments to each dataset. The bag fraction is the percentage of data that is randomly
selected to fit each tree. The learning rate is a measure of how much each individual tree influences the
model fitting process; a lower learning rate means that each tree contributes less to the overall model,
and more trees are fitted (Elith et al., 2008). Tree complexity sets the number of nodes in the tree;
higher tree complexity generally results in fewer, but more complex, trees (Elith et al., 2008).
Tree-based models have been successfully used to characterize and predict a range of ecological and
environmental data (see Cutler et al., 2007; De'ath, 2007; De'ath and Fabricius, 2000; Elith et al., 2008;
and Maloney et al., 2009). In addition, modern tree-based approaches such as BRT have been designed
to improve predictive performance and address the bias issues that sometime arise in tree-based
models (Elith et al., 2008).
All water quality and biological BRT models were fit in the statistical software program R using packages
{dismo} and {gbm} (Hijmans et al. 2015; Ridgeway 2015). Additional information about R can be found at
the Comprehensive R Archive Network (CRAN, https://www.cran.r-project.org/).
D.3 Metrics Methods
Response Variable Management
The following filters were applied to the water quality response variables (total nitrogen [TN], total
phosphorus [TP], and specific conductance [SC]):
• The Activity Category field in the Tennessee Department of Environment and
Conservation (TDEC) water quality (WQ) database was restricted to 'routine samples.'
• Only river/stream sample sites in the Primary Type field were retained.
• A sample date threshold of 1/1/2000 was applied.
Tennessee Integrated Assessment of Watershed Health
D-2
-------
• Observations with quality codes of NULL, below the laboratory detection limit, or
between laboratory detection and quantification limit were kept; all other quality codes
were discarded.
• Observations with values below the laboratory detection limit or values between the
laboratory detection and quantification limit were replaced with values half the
detection or quantification limit, respectively.
• Only NHDPIus catchments with five or more post-2000 samples were retained.
• The median parameter value within each NHDPIus catchment was calculated.
• A power transformation was applied as needed to make each response variable
approximately normal.
Similar filters were applied to the response variables to assess Habitat Condition and Biologic Condition
taken from the TDEC database (Rapid Bioassessment Protocol Score, Benthic BioRecon). Because the
nature of the habitat and biology data are different from water quality, and there were fewer samples
from which to choose, the filters were, by nature, not as strict:
• Only river/stream sample sites in the Primary Type field were retained.
• A sample date threshold of 1/1/2000 was applied.
• The average parameter value in each NHDPIus catchment was calculated.
• A power transformation was applied as needed to make each response variable
approximately normal.
• The largest 5% of NHDPIus catchments (based on cumulative drainage area), by
ecoregion, was excluded from the analysis because TDEC sampling protocols for benthos
and habitat do not include large rivers.
An even smaller sample size for fish Index of Biological Integrity (IBI) values from the Tennessee Valley
Authority (TVA) and Tennessee Wildlife Resources Agency (TWRA) meant that only the "most recent"
sample from each site was used, regardless of date; yet the average parameter value in each NHDPIus
catchment was calculated and a power transformation was applied if necessary.
D.4 Model Fitting
Initial BRT models were fit to each WQ parameter using all available predictor variables. The
appropriateness of the BRT approach for these data was evaluated by examining the following outputs:
• A plot of fitted versus observed values.
• The overall correlation between fitted and observed values for observations used in the
fitted model.
Tennessee Integrated Assessment of Watershed Health
D-3
-------
• The mean cross-validated correlation, which is calculated within each data subset, or
fold, using predicted values for observations that were not used in the fitted model.
• A histogram and density plot of model residuals.
The relative influence value for each predictor variable was then examined. The relative influence of a
predictor represents how many times the variable is selected to split a dataset, weighted by the overall
improvement in model performance across all fitted trees (Elith et al., 2008). Predictors with lower
relative influence scores are not necessarily unimportant; rather, these variables are estimated to have
less influence on the response (Elith et al., 2008). All predictor variables with an initial relative influence
greater than or equal to 0.5 were retained for further analysis. A Spearman correlation matrix was then
computed for this subset of predictor variables and correlation values greater than 0.8 were examined.
In cases where highly correlated variables were likely to contain similar information over the same
spatial scale (i.e., percent catchment Hydrologically Active Zone [HAZ] high intensity development and
percent catchment HAZ impervious cover), one predictor was dropped. However, correlated predictors
were retained if the spatial or temporal processes represented by the variables might reasonably be
expected to influence the response variable in different ways; for instance, the impact of cumulative
riparian forest cover (i.e., natural riparian buffer) is likely to be different than cumulative watershed
forest cover (which may or may not be near riparian areas).
The resulting subset of predictor variables was then used to fit a final model for each WQ parameter.
Per Elith et al. (2008), a range of BRT settings was evaluated. Bag fraction was tested at values of 0.5,
0.6, and 0.75. Learning rate was tested at 0.005, 0.0075, and 0.01. Tree complexity was evaluated at
values of 2, 3, 4, and 5. Model settings were evaluated in terms of cross-validated mean correlation and
the number of trees produced by the final model. Because BRT models select subsets of data
stochastically (i.e., randomly), evaluating the fit of any single BRT output provides a limited perspective
on model performance; fitting the same BRT model two times will produce (slightly) different results. To
better characterize BRT performance, an iterative statistical simulation was applied to the model fitting
process. Each finalized model was fit to the data 100 times. For each iteration, the training and cross-
validated correlation values were retained. In addition, a third goodness of fit (GOF) metric, a
training/test correlation, was also calculated. A training/test correlation is a measure of predictive
performance, similar to the cross-validated correlation but considered to be more indicative of overall
model predictive power (Edith et al., 2008). The training dataset is created by randomly selecting 80% of
the data; the BRT model is fit to these data. The resulting model is then used to predict WQ parameter
values in the test dataset; for example, the remaining 20% of the data that were not used in the model
fitting process. This simulation produced a distribution of GOF statistics, from which mean, median, and
other percentile values could be extracted to evaluate overall model performance.
The finalized BRT model was then used to make WQ parameter predictions for all NHDPIus catchments
in Tennessee. Because habitat and biology data are very much tied to ecoregional differences, separate
models were run, where possible, for the following combinations of Level III ecoregions (in some cases,
ecoregions were combined to increase the sample size):
Tennessee Integrated Assessment of Watershed Health
D-4
-------
Blue Ridge
• Ridge and Valley
• Southwestern and Central Appalachians
• Interior Plateau
• Southeastern Plains, Mississippi River Alluvial Plain, and Mississippi Valley Loess Plains
D.5 Water Quality Model Results
Median Stream Total Nitrogen Concentrations (mg/L)
Total nitrogen in milligrams per liter (mg/L) was calculated by summing nitrate, nitrite, and total Kjeldahl
nitrogen. A power transformation of 0.16 was applied to make the distribution of TN approximately
normal. Table D-l details sample size and summary statistics for median catchment TN.
Table D-l. Summary statistics (minimum, 1st quartile, median, mean, 3rd quartile, and maximum
values) and sample count for median total nitrogen (mg/L) in catchments with five or
more post-2000 samples.
Count
Minimum
Q1
Median
Mean
Q3
Maximum
1,690
0.0735
0.405
0.7243
0.9177
1.184
8.57
Model results indicate a strong regional influence on TN concentrations (Figure D-2). Total forest cover
and agriculture, at both the cumulative and HAZ spatial scales, are also important. Surface and bedrock
geology and developed areas round out the list of important predictors.
Goodness of fit results for the total nitrogen model are acceptable (Figure D-3 and Table D-2). The high
training data correlation values relative to the cross-validated values indicate some overfitting in the
modeling process. However, the cross-validated estimates of predictive power fall within the range of
the training/test results, which indicates that the cross-validated estimate is a relatively unbiased
predictor of predictive performance. The interquartile range (IQR) of the training/test results is narrow
(0.77-0.80), which indicates that predictive performance is relatively stable (Table D-2).
Tennessee Integrated Assessment of Watershed Health
D-5
-------
Figure D-2. Relative influence of predictors in TN model. Note: 'C' indicates cumulative value and
HAZ indicates the hydrologically active zone.
Level 4 EcoRegion
HAZ Total Forest, C
Total Forest, C
Dominant Bedrock (Rock Type 1), C
Agriculture, C
HAZ Agriculture, C
HAZ Developed Open Space, C
HAZ Open Water, C
Developed High Intensity, C
Developed Low Intensity, C
Dominant Surface Geology, C
Dominant Bedrock (Rock Type 1)
5 10 15 20
Relative Influence
~
~
r
Tennessee Integrated Assessment of Watershed Health
D-6
-------
Figure D-3. Simulated distributions of goodness of fit metrics for the TIM model.
0 0 0.
00
0.1
02
Training Data Correlation
12 o
Cross-Validated Mean Correlation
I * I I
00 0 1 02 03
Training/Test Prediction Correlation
i i i i
03 0.4 05 0.6
Correlation
0 4 0 5 0.6 0 7 0 8 0.9 1 0
I I I I I I I
04 05 06 07 0.8 0.9 1 0
0 9
i
1 0
Table D-2. Summary statistics for simulated goodness of fit metrics for median TN model.
Statistic
Minimum
Q1
Median
Mean
Q3
Maximum
Training Correlation
0.84
0.87
0.88
0.88
0.89
0.91
Cross-Validated Mean Correlation
0.77
0.78
0.79
0.79
0.79
0.80
Training/Test Correlation
0.73
0.77
0.79
0.79
0.80
0.84
Tennessee Integrated Assessment of Watershed Health
D-7
-------
Median Stream Total Phosphorus Concentrations (mg/L)
A log transformation was applied to median TP concentrations to make the distribution approximately
normal. Table D-3 details sample size and summary statistics for median catchment TP.
Table D-3. Summary statistics (minimum, 1st quartile, median, mean, 3rd quartile, and maximum
values) and sample count for median TP (mg/L) in catchments with five or more post-2000
samples.
Count
Minimum
Q1
Median
Mean
Q3
Maximum
1,828
0.002
0.01
0.02088
0.07135
0.095
0.9
Model results indicate a very strong regional influence, followed by dominant bedrock (Figure D-4).
Geology is a strong influence on TP values in the state, with geologic weathering contributing phosphate
in several regions (USGS, 1967). Geologic factors appear to be more influential on TP at the statewide
scale than land use (Figure D-4). However, the percent forest in the HAZ is also important; riparian forest
cover can act as a buffer for phosphorus runoff from agricultural and developed land uses. In addition,
row crops and developed land uses at various intensities and spatial scales were also influential.
Tennessee Integrated Assessment of Watershed Health
D-8
-------
Figure D-4. Relative influence of predictors in TP model. Note: 'C' indicates cumulative value and
HAZ indicates the hydrologically active zone.
Level 4 EcoRegion
Dominant Bedrock (Rock Type 1), C
HAZ Total Forest, C
Dominant Bedrock (RockTypel)
HAZ Developed High Intensity, C
Channel Length, C
HAZ Cultivated Crops, C
Dominant Bedrock (RockType2)
HAZ Developed Medium Intensity, C
10 20 30
Relative Influence
I
]
]
]
]
]
]
Goodness of fit metrics for the median TP model indicate a good predictive fit (Figure D-5; Table D-4).
The distribution of training data statistics reveals some overfitting, but the cross-validated and
training/test results are aligned. Compared to the TN model, the IQR of the training/test metric is larger,
which indicates slightly less stable predictive performance.
Tennessee Integrated Assessment of Watershed Health
D-9
-------
Figure D-5. Simulated distributions of goodness of fit metrics for the TP model.
Training Data Correlation
I 8 I I i i If i i 1 I
0.0 0 1 0.2 0 3 0 4 0 5 0 6 0 7 0 8 0.9 10
Cross-Validated Mean Correlation
i r i i i i j i r i i
0.0 0 1 0.2 03 04 05 06 07 0 8 0 9 1 0
Training/Test Prediction Correlation
r i i r I i iiii i
00 0 1 02 03 04 05 06 0.7 08 0 9 1 0
Correlation
Table D-4. Summary statistics for simulated goodness of fit metrics for the median TP model.
Statistic
Minimum
Q1
Median
Mean
Q3
Maximum
Training Correlation
0.86
0.88
0.88
0.88
0.89
0.90
Cross-Validated Mean Correlation
0.78
0.79
0.79
0.79
0.80
0.81
Training/Test Correlation
0.71
0.78
0.79
0.79
0.81
0.84
Tennessee Integrated Assessment of Watershed Health
D-10
-------
TN:TP Ratio (Not Used in Assessment)
In addition to predicted median IN and TP concentrations, the TN:TP ratio was calculated by dividing
predicted TN by predicted TP on a catchment by catchment scale (Figure D-6). High TN:TP ratio values
may indicate relatively undisturbed systems, since natural sources of nitrogen tend to be larger than
natural sources of phosphorus. However, considerable natural variation can occur, even in low-nutrient
oligotrophic water bodies (Bergstrom, 2010). Similarly, a low ratio value may indicate eutrophic
conditions due to either natural or anthropogenic causes. Ratio values in lotic systems may also be
highly determined by hydrologic regime (Green and Finlay, 2010). These uncertainties make ranking
TN:TP ratio values difficult; a high ratio value may or may not correspond to a 'healthy' condition. For
these reasons, ratio values were not incorporated into the Water Quality Sub-Index. Results are
presented in the appendix since the information may be useful from a management perspective.
Figure D-6. Spatial distribution of predicted TN:TP ratios.
Median Specific Conductance (nS/cm)
A power transformation of 0.6 was applied to SC values. Sample size and summary statistics for SC are
listed in Table D-5.
Table D-5. Summary statistics (minimum, 1st quartile, median, mean, 3rd quartile, and maximum
values) and sample count for median SC (nS/cm) in catchments with five or more post-
2000 samples.
Count
Minimum
Q1
Median
Mean
Q3
Maximum
1,677
8.25
121
252.5
263.4
385.5
1,121
The SC model results again indicate a strong regional influence, followed by bedrock and surface geology
(Figure D-7). Conductivity is highly influenced by geology; soil and bedrock materials vary tremendously
Tennessee Integrated Assessment of Watershed Health
D-ll
-------
in terms of their ability to ionize. SC is also correlated with dissolved solids and land uses and covers
prone to erosion and sediment loss.
Figure D-7. Relative influence of predictors in SC model. Note: 'C' indicates cumulative value and
HAZ indicates hydrologically active zone.
~
~
n
n
i
Level 4 EcoRegion
Dominant Bedrock (Rock Type 1), C
Dominant Surface Geology, C
HAZ Total Forest, C
HAZ Urban, C
Shrub/Scrub, C
Forest, C
Mean Elevation, C
Dominant Bedrock (Rock Type 2)
Developed Open Space, C
Grassland/Herbaceous, C
0 10 20 30 40
Relative Influence
The goodness of fit metrics for SC indicate very good predictive performance (Figure D-8; Table D-7).
While overfitting is again indicated, it is less apparent than in the other WQ parameter models, and the
cross-validated and training/test metrics are significantly higher. The training/test IQR is also narrow
(0.89-0.91), which indicates stable predictive performance.
Tennessee Integrated Assessment of Watershed Health
D-12
-------
Figure D-8. Simulated distributions of goodness of fit metrics for the SC model.
Training Data Correlation
ll
u
I P I i I I R I 1 I
0.0 0 1 0 2 0 3 0 4 0.5 0.6 0 7 0 8 0.9 1.0
Cross-Validated Mean Correlation
s
III!
j.
*1
.1
i
0 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0
Training/Test Prediction Correlation
ii i i r i i i i i i
0 0 0 1 0 2 0 3 0 4 0 5 0 6 0.7 0 8 0 9 1.0
Correlation
Table D-6. Summary statistics for simulated goodness of fit metrics for median SC model.
Statistic
Minimum
Q1
Median
Mean
Q3
Maximum
Training Correlation
0.95
0.97
0.97
0.97
0.97
0.98
Cross-Validated Mean Correlation
0.88
0.89
0.90
0.90
0.90
0.91
Training/Test Correlation
0.85
0.89
0.90
0.90
0.91
0.93
Tennessee Integrated Assessment of Watershed Health
D-13
-------
D.6 Habitat Model Results
Rapid Bioassessment Protocol Scores
Separate models were run for each ecoregion or group of ecoregions for the Rapid Bioassessment
Protocol (RBP) score. RBP scores range from 0-200. Table D-7 details sample size and summary statistics
for the RBP score.
Table D-7. Summary statistics (minimum, 1st quartile, median, mean, 3rd quartile, and maximum
values) and sample count for the RBP score, by ecoregion.
Ecoregion
Count
Minimum
Q1
Median
Mean
Q3
Maximum
Blue Ridge
284
73
132
152
150.43
172
196
Ridge and Valley
939
38
106
126
124.59
145
184.5
Central/SE Appalachians
473
40
132
151
147.72
166.75
200
Interior Plateau
1,597
52
118
130
130.07
143.5
188
SE/Mississippi Loess Plains
877
31
81
97.5
100.20
115.27
191
Model results vary by ecoregion, but indicate a strong influence of the land use in the HAZ on the RBP
metric (Figure D-9). Distance to bedrock is the strongest driver for the model in the Interior Plateau
region of the state.
The goodness of fit metrics indicate very good predictive performance in all ecoregions (Table D-8).
Tennessee Integrated Assessment of Watershed Health
D-14
-------
Figure D-9. Relative influence of predictors for RBP. Note: "C" indicates a cumulative value and HAZ
indicates hydrologically active zone.
Blue Ridge
Total Forest, C
Total Drainage Area, C
Percent Impervious Surface
Local Total Forest
HAZ Total Forest, C
HAZ Grassland/Herbaceous, C
Dominant Bedrock Upstream
Relative Influence
Ridge and Valley
Total Drainage Area, C
Mean Elevation, C
HAZ Total Forest, C
HAZ Total Forest
HAZ Natural Lands
Dominant Upstream Bedrock
Depth to Bedrock
Relative Influence
(continued)
Tennessee Integrated Assessment of Watershed Health
D-15
-------
Figure D-9. Relative influence of predictors for RBP. Note: "C" indicates a cumulative value and
HAZ indicates hydrologically active zone, (continued)
Central/SE Appalachians
Total Drainage Area. C
Stream Length. C
Percent Impervious Surface
HAZ Total Forest C
HAZ Total Forest
HAZ Total Agriculture
HAZ Grassland/Herbaceous
HAZ Evergreen Forest. C
Developed, Open Space. C
Relative Influence
Interior Plateau
Total Drainage Area. C
Stream Length. C
Mean Elevation. C
HAZ Total Forest
HAZ Natural Lands. C
HAZ Natural Lands
HAZ Deciduous Forest C
HAZ Deciduous Forest
Dominant Bedrock Upstream
Relative Influence
(continued)
Tennessee Integrated Assessment of Watershed Health
D-16
-------
Figure D-9. Relative influence of predictors for RBP. Note: "C" indicates a cumulative value and
HAZ indicates hydrologically active zone, (continued)
SE/Mississippi Loess Plains
Mean Elevation. C
Local Cultivated Crops. C
Local Cultivated Crops
HAZ Total Agriculture. C
HAZ Natural Lands
HAZ Cultivated Crops. C
HAZ Cultivated Crops
Developed Open Space
Relative Influence
Tennessee Integrated Assessment of Watershed Health
D-17
-------
Table D-8. Summary statistics for simulated goodness of fit metrics for RBP model.
Ecoregion
Minimum
Q1
Median
Mean
Q3
Maximum
Ridge and Valley
Training Correlation
0.74
0.78
0.81
0.80
0.84
0.84
Cross-Validated Mean Correlation
0.52
0.54
0.55
0.55
0.57
0.57
Training/Test Correlation
0.47
0.52
0.53
0.54
0.56
0.61
Blue Ridge
Training Correlation
0.83
0.85
0.86
0.86
0.88
0.90
Cross-Validated Mean Correlation
0.54
0.58
0.59
0.59
0.61
0.64
Training/Test Correlation
0.48
0.57
0.58
0.59
0.65
0.68
Central/SE Appalachians
Training Correlation
0.71
0.73
0.75
0.75
0.78
0.79
Cross-Validated Mean Correlation
0.44
0.50
0.51
0.51
0.54
0.55
Training/Test Correlation
0.42
0.45
0.49
0.50
0.58
0.61
SE/Mississippi Loess Plains
Training Correlation
0.79
0.79
0.80
0.81
0.83
0.86
Cross-Validated Mean Correlation
0.67
0.69
0.69
0.69
0.70
0.71
Training/Test Correlation
0.65
0.67
0.68
0.69
0.70
0.75
Interior Plateau
Training Correlation
0.71
0.73
0.75
0.75
0.76
0.80
Cross-Validated Mean Correlation
0.51
0.53
0.55
0.54
0.55
0.57
Training/Test Correlation
0.47
0.54
0.56
0.56
0.58
0.64
Habitat Suitability Ranking
Data provided by The Nature Conservancy (TNC)/TWRA already included a prioritization of NHDPIus
catchments in Tennessee from a low probability of encountering greatest conservation need (GCN)
species to a very high probability. For the purposes of this Assessment, the ranking for only the High and
Very High catchments was used in order to increase the Habitat Condition Sub-Index scores for those
watersheds.
D.7 Biology Model Results
Benthic Macroinvertebrate Score
Separate models were run for each ecoregion or group of ecoregions for the Benthic BioRecon score.
Table D-9 details sample size and summary statistics for the Benthic BioRecon score.
Tennessee Integrated Assessment of Watershed Health
D-18
-------
Table D-9. Summary statistics (minimum, 1st quartile, median, mean, 3rd quartile, and maximum
values) and sample count for the Benthic Macroinvertebrate Metric, by ecoregion.
Ecoregion
COMID
Count
Minimum
Q1
Median
Mean
Q3
Maximum
Blue Ridge
197
3
9
11
10.65
14
15
Ridge and Valley
586
3
6
9
9.49
13
15
Central/SE Appalachians
250
3
9
12
10.98
14.33
15
Interior Plateau
1,289
3
9.67
13
11.75
15
15
SE/Mississippi Loess Plains
648
3
6
10
9.78
13
15
Model results vary by ecoregion, but indicate a strong influence of cumulative upstream land use on the
benthic macroinvertebrate community (Figure D-10). Stream size and elevation also influence the
biology of the watershed, especially in the more mountainous Appalachian and Blue Ridge ecoregion.
The goodness of fit metrics indicate relatively good predictive performance in most ecoregions
(Table D-10). Performance was best in the Blue Ridge and Central/SE Appalachians. The weakest
performing model was in the Ridge and Valley ecoregion.
Tennessee Integrated Assessment of Watershed Health
D-19
-------
Figure D-10. Relative influence of predictors for the benthic macroinvertebrate scores. Note: "C"
indicates a Cumulative value and HAZ indicates hydrologically active zone.
Blue Ridge
Surface Geology, C
Stream Length. C
Shrub and Scrub, C
Natural Lands, C
Dominant Bedrock Upstream
Bedrock Geology 2
Bedrock Geology 1
Relative Influence
Ridge and Valley
Total Square Drainage Area, C
Percent Forest
HAZ Total Forest, C
Deciduous Forest C
Deciduous Forest
Relative Influence
(continued)
Tennessee Integrated Assessment of Watershed Health
D-20
-------
Figure D-10. Relative influence of predictors for benthic macroinvertebrate scores. Note: "C"
indicates a Cumulative value and HAZ indicates hydrologically active zone,
(continued)
Central/SE Appalachians
Total Square Drainage Area, C
Mean Elevation
HAZ Barren Lands, C
Developed Open Space
Relative Influence
Interior Plateau
Mean Elevation, C
Low Intensity Developed, C
HAZ Cultivated Crops, C
Dominant Bedrock Upstream
Developed Open Space, C
Developed Low Intensity, C
Cultivated Crops, C
Relative Influence
(continued)
Tennessee Integrated Assessment of Watershed Health
D-21
-------
Relative influence of predictors for benthic macroinvertebrate scores. Note: "C"
indicates a Cumulative value and HAZ indicates hydrologically active zone,
(continued)
SE/Mississippi Loess Plains
Total Forest C
Pasture Hay. C
Natural Lands. C
Min Elevation
Mean Elevation. C
HAZ Urban. C
HAZ Total Forest C
HAZ Pasture Hay. C
HAZ Low Intensity Developed Lands
HAZ Cultivated Crops. C
Forested Lands. C
D-22
Figure D-10.
~
~
10 20 30
Relative Influence
Tennessee Integrated Assessment of Watershed Health
-------
Table D-10. Summary statistics for simulated goodness of fit metrics for benthic macroinvertebrate
model.
Ecoregion
Minimum
Q1
Median
Mean
Q3
Maximum
Ridge and Valley
Training Correlation
0.63
0.65
0.69
0.68
0.71
0.72
Cross-Validated Mean Correlation
0.42
0.45
0.47
0.47
0.49
0.51
Training/Test Correlation
0.31
0.42
0.48
0.47
0.54
0.58
Blue Ridge
Training Correlation
0.77
0.82
0.88
0.87
0.90
0.98
Cross-Validated Mean Correlation
0.36
0.42
0.47
0.47
0.52
0.57
Training/Test Correlation
0.29
0.36
0.42
0.45
0.50
0.76
Central/SE Appalachians
Training Correlation
0.80
0.82
0.83
0.84
0.85
0.97
Cross-Validated Mean Correlation
0.39
0.43
0.46
0.46
0.48
0.56
Training/Test Correlation
0.17
0.36
0.47
0.44
0.51
0.61
SE Plains/Mississippi Loess Plains
Training Correlation
0.76
0.78
0.80
0.81
0.83
0.88
Cross-Validated Mean Correlation
0.63
0.65
0.67
0.67
0.68
0.68
Training/Test Correlation
0.56
0.59
0.64
0.65
0.69
0.75
Interior Plateau
Training Correlation
0.73
0.75
0.77
0.77
0.79
0.81
Cross-Validated Mean Correlation
0.57
0.59
0.60
0.60
0.61
0.63
Training/Test Correlation
0.55
0.57
0.62
0.61
0.65
0.73
Fish IBI
Fish I Bis were only available for a limited portion of the state, so the models were not run separately by
ecoregion. Table D-ll details sample size and summary statistics for the fish IBI.
Table D-ll Summary statistics (minimum, 1st quartile, median, mean, 3rd quartile, and maximum
values) and sample count for the fish IBI score.
COMID Count
Minimum
Q1
Median
Mean
Q3
Maximum
102
18
33.5
38
38.92
46
56
Model results show a strong influence of cumulative land use, especially in the HAZ, on the quality of the
fish community in Tennessee (Figure D-ll).
The goodness of fit metrics indicate very good predictive performance (Table D-12).
Tennessee Integrated Assessment of Watershed Health
D-23
-------
Figure D-ll. Relative influence of predictors for fish IBI. Note: "C" indicates a cumulative value and
HAZ indicates hydrologically active zone.
Fish IBI
HAZ C |
HAZ Scrub Shrub, C ]
HAZ Natural C |
Developed High Intensity, C ~|
HAZ Total Forest, C
HAZ Evergreen Forest, C |
Stream Order
Type 2
Total Drainage Area |
Sinuosity |
0 2 4 6 8 10 12
Relative Influence
Table D-12. Summary statistics for simulated goodness of fit metrics for the Fish IBI.
Statistic
Minimum
Q1
Median
Mean
Q3
Maximum
Training Correlation
0.86
0.86
0.87
0.89
0.94
0.99
Cross-Validated Mean Correlation
0.38
0.41
0.44
0.46
0.52
0.56
Training/Test Correlation
0.35
0.39
0.58
0.54
0.67
0.69
Tennessee Integrated Assessment of Watershed Health
D-24
-------
D.8 References
Bergstrom, A-K. 2010. The use of TN:TP and DIN:TP ratios as indicators for phytoplankton nutrient
limitation in oligotrophic lakes affected by N deposition. Aquatic Sciences 72(3), 277-281.
Cutler, D., T.C. Edwards Jr., K.H. Beard, A. Cutler, K.T. Hess, J. Gibson, and J.J. Lawler. 2007. Random
forests for classification in ecology. Ecology 88: 2783-2792.
De'ath, G. 2007. Boosted trees for ecological modeling and prediction. Ecology 88(1): 243-251.
De'ath, G., and K. Fabricius. 2000. Classification and regression trees: A powerful yet simple technique
for ecological data analysis. Ecology 88(11): 3178-3192.
Elith, J., J.R. Leathwick, and T. Hastle. 2008. A working guide to boosted regression trees. Journal of
Animal Ecology 77: 802-813.
Green, M., and J. Finlay. 2010. Patterns of hydrologic control over stream water total nitrogen to total
phosphorus ratios. Biogeochemistry 99, 15-30.
Hijmans, R., S. Phillips, J. Leathwick, and J. Elith. 2015. dismo: Species Distribution Modeling. R package
version 1.0-12. Available at http://CRAN.R-proiect.org/package=dismo
Maloney, K., D. Weller, M. Russel, and T. Hothorn. 2009. Classifying the biological condition of small
streams: an example using benthic macroinvertebrates. Journal of the North American
Benthological Society 28(4): 869-884.
Ridgeway, G. 2015. gbm: Generalized Boosted Regression Models. R package version 2.1.1.
http://CRAN.R-proiect.org/package=gbm.
United States Geological Survey (USGS). 1967. Phosphate deposits: A summary of salient features of the
geology of phosphate deposits, their origin, and distribution. Geological Survey Bulletin 1252-D.
Available at http://pubs.usgs.gov/bul/1252d/report.pdf
Tennessee Integrated Assessment of Watershed Health
D-25
-------
Appendix E: Data Analyses Methods and Correlation Results
E.l. Background
Data transformations, including normalization, standardization, centering, and scaling, are often
required to complete analyses of environmental and ecological data. This is because the results may be
biased if data are not transformed prior to analysis. However, the selection of the transformation
method is guided by project goals and end user considerations, such as:
• Need to combine/integrate multiple variables that occur on different scales so that
results are not biased towards any single attribute.
• Need to directly compare catchments on a relative basis.
• Need to provide the end-user with a useable tool that is amenable to further
development in a format that does not report too many 'significant digits.'
Based on these goals and considerations, rank normalization was selected as the transformation method
for this Assessment. Rank normalization transforms one or more variables to a uniform distribution and
scale, typically from 0 to 100; this common scale allows for comparisons between variables that may
exhibit different units and scales. Rank normalization is also insensitive to outlier or extreme values,
which can overly compress a normalized distribution when other normalization methodologies are
applied (Mitchell, 2012).
However, the effects of standardizing the scale and distributing component metrics are not always
positive, particularly when the values of a metric are predominantly in a range considered to be "good"
or predominantly "poor." It is also important that rank-normalized scores with lower index and sub-
index scores should not be considered impaired or degraded; rather, the condition is lower in score
relative to other catchments in the Assessment area. If all the catchments in a basin are considered
"good" for a given metric, catchments with the lower metric scores will be considered the "least"
healthy. Rank normalization can also be problematic when a large number of catchments share the
same value of a given metric. The risk of these undesirable outcomes was minimized by choosing
component parameters in consultation with the Tennessee Healthy Watershed Initiative Technical Team
as well as examining the observed variability of candidate variables; if a parameter was not judged to be
indicative of watershed health or vulnerability or exhibited very low variability, the variable was not
included in the Assessment.
E.2. Rank Normalization of Metrics
Metrics of watershed health and vulnerability were rank normalized for reporting the metric, sub-index,
and final index calculations. Rank normalization provides metric scores ranging from 0 to 100 with
consistent directionality. Rank normalizing the watershed health metrics involved the following steps:
• Rank all catchments on the basis of raw metric scores:
Tennessee Integrated Assessment of Watershed Health
E-l
-------
- Catchments were ranked in ascending order if higher metric scores corresponded to higher
watershed health (i.e., higher percent natural lands).
- Catchments were ranked in descending order if lower metric scores corresponded to higher
watershed health (i.e., low total nitrogen concentrations).
• Apply the following formula to calculate the catchment's rank-normalized score:
Catchment Rank — Minimum Rank
Rank Normalization = x 100
Maximum Rank — Minimum Rank
For this Assessment, the minimum rank was always 1 and the maximum rank was 61,859 (the total
number of NHDPIus catchments in Tennessee). The catchment rank was based on the order of the raw
metric scores.
Rank-normalized scores are directionally aligned so that higher scores for watershed health metrics and
sub-indices correspond to higher watershed health (Table E-l). The results of each metric and sub-index
are displayed in the Map Atlas (Appendix A) using colors to depict the final score of each catchment;
cool (blue) colors represent better condition and warm (yellow) colors represent lower condition.
As noted above, rank normalization was not applied to the Hydrologic Condition metrics. Instead, only
the final Hydrologic Condition Sub-Index was rank normalized based on the sum of the absolute
percentage change for all the component metrics. Lower scores correspond to the largest total
percentage change, and higher scores correspond to the smallest total percentage change across the
watershed.
E.3 Multi-metric Index Development
Multi-metric indices are a powerful tool for reporting aggregate conditions for ecosystems, including
healthy watersheds. At the same time, care is required to ensure that multi-metric indices remain
transparent and are not confounded with redundant or spurious information.
Index scores were aggregated at two levels: the sub-indices (six for watershed health and three for
vulnerability) and the Watershed Health and Vulnerability Indices. Metrics were first combined into a set
of sub-indices based on the groupings depicted in Figure 7. Each sub-index describes one attribute or
component of watershed health (Landscape Condition Sub-Index, Geomorphic Condition Sub-Index,
Hydrologic Condition Sub-Index, Water Quality Sub-Index, Habitat Condition Sub-Index and Biological
Condition Sub-Index) or vulnerability (Land Use Vulnerability Sub-Index, Water Vulnerability Sub-Index,
and Climate Change Vulnerability Sub-Index). The purpose of scoring the sub-indices before calculating
the Watershed Health and Vulnerability Indices was to balance the influence of each metric on the
overall index scores. Without this step, index scores could be biased toward attributes with the higher
number of metrics (e.g., Hydrologic Condition).
Tennessee Integrated Assessment of Watershed Health
E-2
-------
Table E-l. Original directionality of watershed health and vulnerability metrics.
Metric
Original Directionality
Watershed Health
Percent Natural Land Cover
Higher values = Higher watershed health
Percent Natural Lands in HAZ
Resistance Score
Rapid Bioassessment Protocol (RBP)
Species Habitat Suitable Score
Benthic Macroinvertebrate Score
Fish IBI
Erosion Score
Lower value = Higher watershed health
Deviation from Streamflow Characteristics
Reference Condition
Dam Storage Ratio
Stream Total Nitrogen Concentration
Stream Total Phosphorus Concentrations
Stream Specific Conductance
Watershed Vulnerability
Projected Impervious Cover Change
Higher value = Higher Watershed
Vulnerability
Potential for Energy Development
Projected Change in Water Consumption
Projected Change in Water Withdrawal
Projected Increase in Drought
Projected Increase in Heavy Precipitation
Events
The rank-normalization methodology provides metric scores that are directionally aligned (i.e., higher
rank-normalized scores correspond to higher watershed health). Index scores follow the same
directionality such that High Watershed Health Index scores correspond to high watershed health and
high Vulnerability Index scores corresponds with watersheds that are the most like to be negatively
impacted by future projected changes in land use, water use, and climate. Rank normalization eased
interpretation by providing scores that correspond to percentiles. For example, a Watershed Health
Index score of 75 corresponds to the 75th percentile of condition.
E.4 Correlation Analyses
A correlation analysis between all possible pairings of the watershed health sub-indices was conducted
to determine whether there is any relationship between these calculated measures that would
ultimately prohibit combining the sub-indices into the Watershed Health Index without redundancy. The
potential for correlation exists due to some commonalities in the underlying data used to calculate the
metrics used for each sub-index. Component metrics for Geomorphic Condition, Hydrologic Condition,
Biological Condition, Habitat Condition, and Water Quality were quantified from statistical models that
relate stream health observations to several landscape variables, including those that describe the
Tennessee Integrated Assessment of Watershed Health
E-3
-------
amount and distribution of natural land cover in a catchment. Because these same properties are
captured in Landscape Condition, there is potential for redundancy between the Landscape Condition
Sub-Index and sub-indices derived from modeled metrics, as well as between the sub-indices
themselves.
Table E-2 presents the Pearson correlation coefficients among each combination of sub-indices. Values
range from 0.008 to 0.485. A weak correlation exists between the Geomorphology and Habitat
Condition Sub-Indices (R2 = 0.485); however, this correlation does not violate the redundancy limits used
in other studies of biological multi-metric index development (Emery et al., 2003; Hering et al., 2006;
Stoddard et al., 2008). These correlation results support the use of all sub-indices in the calculation of
the Watershed Health Index without concerns of redundancy.
Table E-2. Pearson correlation coefficients (R-squared values) between each pairing of sub-indices
available to calculate the Watershed Health Index.
Metric
Geomorphology
Hydrology
Water Quality
Habitat
Biology
Landscape
0.287
0.008
0.362
0.253
0.157
Geomorphology
0.044
0.267
0.485
0.088
Hydrology
0.016
0.040
0.039
Water Quality
0.217
0.226
Habitat
0.146
E.5 References
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(4): 791-808.
Hering, D., C.K. Feld, O. Moog, and T. Ofenbock. (2006). Cook book for the development of a Multimetric
Index for biological condition of aquatic ecosystems: experiences from the European AQEM and
STAR projects and related initiatives. Hydrobiologia 566(1): 311-324.
Mitchell, H.B. 2012. Data Fusion: Concepts and Ideas. Springer, 2nd edition. 346 pages.
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
Benthological Society 27(4): 878-891.
Tennessee Integrated Assessment of Watershed Health
E-4
------- |