United States
Environmental Protection
Agency
Assessment of Wetland Ecosystem Condition
across Landscape Regions:
A Multi-metric Approach
Part A. Ecological Integrity Assessment Overview
and Field Study in Michigan and Indiana
EPA/600/R-12/021a
June 2012
www.epa.gov
-------
EPA/600/R-12/021a
June 2012
www.epa.gov
Assessment of Wetland Ecosystem Condition across
Landscape Regions: A Multi-metric Approach
Part A. Ecological Integrity Assessment Overview
and Field Study in Michigan and Indiana
Don Faber-Langendoen1, Cloyce Hedge2, Mike Kost3, Steve Thomas3, Lindsey Smart1, R. Smyth1
Jim Drake1and Shannon Menard1
^atureServe, Conservation Science Division, 4600 N. Fairfax Dr., 7th floor, Arlington, VA 22203
2 Indiana Natural Heritage Program, Division of Nature Preserves, Department of Natural Resources.
402 West Washington, Rm. W267, Indianapolis, IN 46204
3 Michigan Natural Features Inventory, Michigan State University-Extension, P.O. Box 30444,
Lansing, Ml 48909
Notice: Although this work was reviewed by EPA and approved for publication, it may not
necessarily reflect official Agency policy. Mention of trade names and commercial products
does not constitute endorsement or recommendation for use.
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
NOTICE
The United States Environmental Protection Agency (EPA) through its Office of
Research and Development funded and managed the research described here via a
grant (#R-83377501). It has been reviewed by the EPA and approved for publication.
Mention of trade names or commercial products does not constitute endorsement
or recommendation by EPA for use.
11
-------
FOREWORD
EPA's Environmental Monitoring and Assessment Program (EMAP) is a research program to
develop the tools necessary to monitor and assess the status and trends of national ecological
resources over broad spatial and temporal scales. Regional EMAP (REMAP) is a partnership
between the EPA Regional Offices and EPA's Office of Research and Development (ORD), with
the goal of building state and tribal capacity for using statistically valid monitoring data for
reporting on the condition of their aquatic resources. ORD works with the Regional Offices to
provide funds for projects meeting EMAP criteria that are of importance to the needs within
the region. In the REMAP 2007 funding announcement, one of the identified priority focus
areas was the "Development and testing of protocols and/or the monitoring and assessment
of wetlands in the Region 5 states using a stratified, statistically-valid sample survey design that
will allow extrapolation of wetland conditions throughout ecological regions of the Midwest".
Under a competitive process, a Cooperative Agreement (R-83377501) was awarded to
NatureServe for the proposal they submitted to this focus area. .
This report describes the results of NatureServe's project "Assessment of Wetland Ecosystem
Conditions across Landscape Regions - a Multi-metric Approach". The project was conducted
in partnership with the Natural Heritage programs of Indiana and Michigan, and included
assessment of ~360 wetland sites in those two states. Main elements of the project include
examining the suitability of existing spatial datasets and classification systems as the basis for
sampling design, developing and assessing metrics for various aspects of wetland condition, and
synthesizing the results into an ecological integrity scoring system.
Anett Trebitz (Mid-Continent Ecology Division, Duluth MN), was the EPA Project Officer,
providing administrative oversight and technical input and reviews. Other individuals at EPA
who provided input or reviews included Sue Elston (Region 5, Chicago IL), Peter Jackson (Region
5, Chicago IL), Mike Scozzafava (Office of Wetlands, Washington DC), and Rich Sumner
(Regional liaison for the National Wetlands Program, Corvallis OR). Jo Thompson (REMAP
Coordinator, Mid-Continent Ecology Division, Duluth MN) facilitated the funding announcement
and selection process and David Ack (Grants Management Division, Washington DC) was the
grant specialist for the project.
EPA's Mid-Continent Ecology Division is publishing this report to make these findings more
widely available, given their potential significance for EPA's new National Wetlands Condition
Assessment, as well as for state or tribal agencies involved in assessments of their wetland
resources.
Carl Richards,
Director,
EPA, Office of Research and Development, Mid-Continent Ecology Division
in
-------
This document is the Irst of a two part publication.
PART A (this publication):
Faber-Langendoen, D., C. Hedge, M. Kost, S. Thomas, L. Smart, R. Smyth, J. Drake, and S.
Menard. 2012a. Assessment of wetland ecosystem condition across landscape regions: A
multi-metric approach. Part A. Ecological Integrity Assessment overview and field study in
Michigan and Indiana. EPA/600/R-12/021a. U.S. Environmental Protection Agency Office of
Research and Development, Washington, DC.
PART B:
Faber-Langendoen, D., J. Rocchio, S. Thomas, M. Kost, C. Hedge, B. Nichols, K. Walz, G. Kittel, S.
Menard, J. Drake, and E. Muldavin. 2012b. Assessment of wetland ecosystem condition
across landscape regions: A multi-metric approach. Part B. Ecological Integrity Assessment
protocols for rapid field methods (12). EPA/600/R-12/021b. U.S. Environmental Protection
Agency Office of Research and Development, Washington, DC.
IV
-------
ACKNOWLEDGEMENTS
Developing the overall methodology for ecological assessments has been a team effort over
many years. The groundwork for this report was provided by the coauthors of Faber-
Langendoen et al. (2008). Their work, in turn, drew from NatureServe's Ecological Integrity
Assessment Working Group. Members consisted of NatureServe and Network member program
staff, including Don Faber-Langendoen and Pat Comer (co-chairs), and David Braun, Elizabeth
Byers, John Christy, Greg Kudray, Gwen Kittel, Shannon Menard, Esteban Muldavin, Milo Pyne,
Carl Nordman Joe Rocchio, Mike Schafale, Lesley Sneddon, and Linda Vance. More recently, the
overall framework for the methodology has been summarized by Unnasch et al. (2009), and we
have drawn from some of that publication for this report.
We are grateful for support from staff of the Environmental Protection Agency, including the
project management provided by Anett Trebitz, and ongoing support from Rich Sumner and
Mike Scozzafava for their support, and for the peer review comments from Rich Sumner,Peter
Jackson and Sue Elston.
The project was undertaken by staff at NatureServe, with field crews coordinated by the
Michigan Natural Features Inventory (MNFI) and the Indiana Natural Heritage Program (INNHP).
We thank our field staff. In Indiana field staff included Roger Hedge, Mike Homoya, Tom Post,
Tom Swinford, Rich Dunbar, Nate Simons, Derek Nimetz, and Jason Larson. Office support,
including compiling site information and data entry, was provided by Ron Hellmich and Breanna
Sowers. In Michigan field staff included John Fody, Aimee Kay, Jeff Lee, Monique Gorecki, Jesse
Lincoln, and Aaron Kortenhoven. Office staff support was provided by Rebecca Rogers, Helen
Enander, Suzanne Ridge, and Nancy Toben.
Critical database support was provided by Kristin Snow and Mary Harkness. They helped design
and implement the Ecological Observations database, which houses all information collected
during the project, and which can be linked to the Biotics databases used by the two Natural
Heritage programs. We thank them for their efforts.
In this document, we provide a full overview of the EIA method along with a field study
in Indiana and Michigan based on the method. A previous version of this study (Faber-
Langendoen et al. 2011) contains the original metric protocols used for the field study.
Our most current metric protocols for Level 2 ElAs of wetlands, based on revisions from
this study, are provided in a companion document (Faber-Langendoen et al. 2012).
-------
TABLE OF CONTENTS
Notice /'/'
Foreword Hi
Acknowledgements v
A.I INTRODUCTION 7
A.2 BACKGROUND 8
Natural Heritage Methodology 8
Project Organization 8
A.3 PURPOSES OF ECOLOGICAL INTEGRITY ASSESSMENTS 9
A.4 THE UNIT OF ASSESSMENT - ECOSYSTEM OBSERVATIONS 10
Ecosystem Types 10
Points, Polygons, and Patches 10
Watersheds & Landscapes 11
A. 5 RANGE OF NA TURAL VARIABILITY AND ECOLOGICAL INTEGRITY 12
The Range of Natural Variability Concept 12
Ecological integrity and Range of Natural Variability 13
Ecological Integrity and Climate Change 14
A. 6 ECOLOGICAL INTEGRITY ASSESSMENT METHOD 15
Purpose of the Assessment 15
Conceptual Model for Terrestrial Ecosystems 15
Indicator Ratings 24
Indicators at Multiple Scales (Level 1 to Level 3) 24
Level 1 Assessment (remote-sensing metrics) 26
Level 2 Assessment (rapid field-based metrics) 32
Level 3 Assessments (intensive field metrics) 35
Field Methods and Protocols 37
Ecological Integrity Scorecards 37
A. 7 DEFINITION OF ECOLOGICAL INTEGRITY RATING VALUES (A-D) 40
Definitions 40
Recap of Natural Variability and Ecological Integrity 40
A.8 RETURNING TO THE WATERSHED OR LANDSCAPE SCALE 42
A.9 ECOLOGICAL INTEGRITY, CONSERVATION STATUS AND ECOSYSTEM
SERVICES. 43
A.JO ADAPTING THE ASSESSMENT OVER TIME 45
B.I INTRODUCTION 46
B.2 SAMPLING DESIGN METHODOLOGY. 47
Project Area 47
Classification 48
Other Classifications 50
Wetland Occurrence Data - Natural Heritage And Other Datasets 52
vi
-------
Sampling Design For Reference Gradient 54
Statistical Tests Of Sampling Design 58
B.3 RESULTS 59
Reference Gradient by Classification Stratum 59
Reference Gradient by Condition Stratum 64
B.4 DISCUSSION. 66
Classification 66
Condition 66
C.I INTRODUCTION 67
C.2 METHODS 67
Level 2: Rapid Field Assessment Methods 67
Level 3: Intensive Field Assessment Methods 72
Data Management 75
Index of Ecological Integrity 76
Statistical Screening of Metrics, Attributes and Index of Ecological Integrity 76
Level 2 and Level 3 metrics (Coefficient of Conservatism) 78
Testing and Applying the Revised EIA Method 78
C.3 RESULTS. 79
Statistical Screening -redundancy 79
Statistical screening - Discriminatory power 81
Final Selection of Metrics 85
Applying the Final Model 87
Comparison with Natural Heritage Ranks 89
Application of IEI scores to wetland sites 90
C.4 DISCUSSION 92
The Ecological Integrity model 92
Overall Level 2 Index and Level 3 metrics 93
Calibrating IEI with remote sensing models 94
Study Design - Assessing Wetlands Versus Points 94
A Standard Method for Assessing Wetland Condition 94
Floristic Quality Assessment (Mean Cn) Michigan and Indiana 119
Level 1 Metrics 725
Level 2 Metrics 126
Level 3 Metrics 727
Level 3 Scorecard. 729
Level 3 Metric Summary by Management Unit 130
vn
-------
EXECUTIVE SUMMARY
General: Many ecosystem monitoring and assessment programs are expanding their focus to address
changes in ecosystem condition. This is a challenging task, given the complexity of ecosystems and the
changes they undergo in response to a variety of human activities and landscape alterations. Agencies
and organizations are in need of ecological methods that can address this aspect of ecosystems. These
methods must be sensitive to variations in ecosystem type, size, landscape setting, along with the
dramatic losses and degradation that have occurred.
NatureServe, in partnership with member programs from the Natural Heritage Network and federal
agencies, has developed an assessment of ecosystems condition, structured around the concept of
ecological integrity. Here we first (section A) review and update the overall conceptual model, and
second (Section B), develop a sampling design for identifying a suite of 277 sites, primarily from Natural
Heritage Program databases. These sites span all wetland type and conditions in southern Michigan and
northern Indiana. Third (Section C), we test the method on the 277 sites to determine whether it could
accurately distinguish ranges of integrity across all wetland types. We tested and applied the method
using a multi-level framework (remote, rapid, intensive) and multiple metrics that cover hydrology, soils,
vegetation, size, buffer, and landscape. Data were summarized using an overall Index of Ecological
Integrity and scorecard, focusing on rapid field assessments scores.
Main Objectives
A. Develop a methodology for assessing wetland condition based on ecological integrity. A
conceptual model was developed to facilitate identification of a scientifically defensible set of
metrics, at multiple scales. The methodology informs two main areas of wetland inventory and
monitoring:
• Baseline inventory and ambient monitoring of wetland condition. Our methodology is applicable
for local, state, and national inventory and monitoring of wetlands over time at a variety of
levels (remote, rapid, intensive).
• Mitigation and restoration of wetlands. Our methodology sets ecological performance standards
to assess site-specific and watershed-scale mitigation and restoration projects. Our methods
complement other rapid assessments that are strictly mitigation focused, such as the Michigan
Rapid Assessment Method (MIRAM). MIRAM provides a rating system that compares a
wetlands functional value (including integrity, ecosystem services, and social value) with other
wetlands in the state, regardless of ecological type.
B. Identify a candidate set of wetlands that span the reference gradient in northern Indiana and
southern Michigan (Omernik level 3 ecoregions 55, 56, and 57). We used an objective screening
process (remote sensing based metrics and previous ground surveys based on Natural Heritage
methodology) to identify candidate sites that spanned the reference gradient (minimally disturbed
to highly degraded). All major wetland types were included. Our remote sensing metrics relied on a
combination of landscape condition and stressor metrics relevant to ecological integrity.
-------
C. Test the effectiveness of the ecological integrity assessment method on the candidate set of
wetland sites and revise the method accordingly. We collected EIA data on 277 sites that comprised
the reference gradient, spanning wetland types and conditions. We statistically tested the metrics,
including use of a Human Stressor Index, to determine which are most helpful in identifying the
range of ecological integrity. We revised the EIA methods and then summarized the ratings for each
metric, ecological attribute, and overall set of rank factors. We used an ecological integrity
"scorecard" to display the metric ratings and generate an overall index of ecological integrity. The
scorecard helps interpret the overall status and trends in ecological integrity at a site and across the
region.
Section A: We developed our wetland ecosystem condition assessment using the concept of ecological
integrity. Building on the related concepts of biological integrity and ecological health, ecological
integrity can be defined as "an assessment of the structure, composition, and function of an ecosystem
as compared to reference ecosystems operating within the bounds of natural or historic disturbance
regimes." To have integrity, an ecosystem should be relatively unimpaired across a range of
characteristics and spatial and temporal scales. This broad definition can serve as a guide to developing
ecological integrity assessment methods that are distinct from related assessment methods for
ecological functions or ecosystem services.
Our multi-metric approach for our Ecological Integrity Assessment (EIA) method is similar to the Index of
Biotic Integrity (IBI) for aquatic systems. Our method builds on the work of other rapid assessment
methods (especially the Ohio Rapid Assessment Method and California Rapid Assessment Method), and
our previous work on standardized methods for assessing ecosystems condition for the Natural Heritage
Network along with setting performance standards for wetland mitigation. Critical to our effort was the
use of conceptual models that highlight ecological factors and attributes for which metrics (or specific
indicators) of integrity are most needed. We defined metrics as values derived from specific measures
(e.g., basal area, stand structural class, species diversity) that inform us about the status of an ecological
factor or attribute of integrity. For our model, the primary rank factors and major ecological factors
were landscape context (landscape, buffer), size, and condition (vegetation, soils, and hydrology). We
then selected key metrics that are most responsive, practical, cost-effective and well-tested in
measuring the condition of the ecosystem. The conceptual model also provided a structure in which to
identify known stressors, or agents of change, that affect these major ecological factors. Together they
can help guide management decisions to maintain or restore ecological integrity.
The EIA method also is expected to function across a wide range of wetland ecosystem types. Itallows
for various levels of assessment (remote sensing and field based, both rapid and intensive sampling
methods), and is structured around the availability of a wide set of indicators and metrics. We provide
an overview and demonstration of these methods, with metrics and scoring at multiple scales of
assessment, and a scorecard summary of the metrics using an index of ecological integrity (IEI).
Section B: The primary focus of this section was to create a sampling design that would allow us to test
the sensitivity of the EIA method to changes in ecological integrity across the full range of wetland types
(e.g., bog, rich fen, marsh, wet meadow, wet prairie, swamp, floodplain forest) and conditions
(minimally disturbed to degraded). Secondarily, if the sampling design was successful, it could also
serve as a screening method for identification of a wetland reference gradient (a set of sites that
represent the range of conditions, from minimally disturbed to degraded). Thus, if we successfully
create a sampling design that hypothetically spans a range of conditions, and is independently verified,
then that design can predict the reference gradient.
-------
Our primary source of potential sites came from the Natural Heritage program databases in Michigan
and Indiana. Both programs have identified high quality (minimally disturbed) "element occurrences"
(EOs) of wetland types across the state. These include locations of wetland types with sufficient size and
condition to have conservation value. Programs use their own state natural community classifications,
which link to the U.S. National Vegetation Classification (USNVC) and NatureServe's Ecological Systems
classification. These reference datasets are unique within these states, as they are for many other
states where Natural Heritage Programs have gathered data on high quality examples of native
ecosystems.
Until recently, Heritage programs typically evaluated the condition or ecological integrity of occurrences
in the field using best professional judgment and with a minimal amount of quantitative information.The
evaluations are summarized using an element occurrence rank (EORANK): A (Excellent), B (Good), C
(Fair), D (Poor). Typically, on-site condition was the primary focus when assigning an EORANK. We
compiled element occurrences (EOs) for several EPA ecoregions of interest across southern Michigan
and northern Indiana. From the available data, we established a site selection process based upon: 1)
wetland type and 2) wetland condition. First, we assigned each wetland element occurrence to the
macrogroup level of the USNVC (seven macrogroups- Bog & Poor Fen, Rich Fen, Wet Prairie, Wet
Shrub, Meadow & Marsh, Coastal Plain Pondshore, Swamp, and Floodplain Forest). This was
straightforward based on the cleanly nested crosswalk of state type to macrogroup type. More
challenging was the condition stratum, where our best source of information is the Heritage EORANK.
To increase the standardization of the EORANK, we combined the Heritage rank (which emphasizes on-
site condition) with a remote-sensing based landscape-context evaluation. This evaluation uses three
primary metrics: naturalness of surrounding landscape, land uses within the landscape, and the extent
and condition of the buffer immediately surrounding the wetland. When combined, the landscape
metrics and Heritage rank create a "condition stratification rating" for each occurrence (landscape
context rating + EORANK rating = condition stratification rating). Minimally disturbed (A and B ranked
sites) are often hard to find in this region because of extensive land conversion to agriculture and other
land uses. Fortunately, Heritage databases tend to emphasize identification of those sites.
To ensure sufficient replication of conditions and wetland types, we developed a pool of 280 possible
sites (7 macrogroups x 4 conditions x 10 replicates). The Heritage databases provided most of the
occurrences needed to fill this design including many minimally disturbed occurrences. But Heritage
databases often lacked degraded occurrences, especially for more common wetland types. We
addressed this by having crews send in possible sites based on drive-bys, and then coupled their
evaluation with the landscape metrics to see if the site qualified as degraded. Field crews tracked their
progress in meeting the sampling design to ensure a representative coverage across the reference
gradient. Over time, crews failed to find some sites and other sites were destroyed or had their original
reported conditions change considerably. In addition some types (Bog & Poor Fen) are relatively rare
and in difficult to access locations, and few degraded examples were available. Our final survey design
had 277 sites.
To test the merits of our sampling design, we tallied the number of sites sampled by macrogroup and
condition. We successfully maintained a balance across the classification stratum - each of the 7
macrogroups had between 30 and 55 survey sites. Thus the predicted wetland type at each site was
typically found during the survey. We were less successful in maintaining a balance of sites across the
full range of condition, partly because our chosen stratification (screening) method under-predicted the
expected number of A condition sites, and over-predicted the number of D condition sites (based on the
outcomes of our field assessment of ecological integrity, summarized in Section C). As a result, we
-------
reexamined the stratification approach, and proposed a modified version that we recommend for future
selection of reference sites.
Thus, our sampling design was sufficiently robust to capture the full range of wetland types and of
conditions (even if not strictly balanced). Second, by redesigning our stratification approach, we are
confident that we can predict site locations for a reference gradient of wetland types and condition
(from minimally disturbed to degraded). Given the widespread availability of both Heritage data and
interpreted remote sensing imagery, we suggest that our methods can be used by studies that need to
identify either benchmark (minimally disturbed) sites, or an entire reference gradient. Knowledge of
these sites is becoming increasingly important, given increasingly degradation and loss of native
ecosystems across many parts of the country.
Section C: In this section we tested and applied our ecological integrity assessment method by
evaluating 277 wetland sites in the field. Our conceptual model provided a framework to identify
metrics for major ecological factors (MEFs), including vegetation, hydrology, soils, size, buffer and
landscape. For the rapid assessment (Level 2 or L2), 18 major metrics across all MEFs were initially used,
along with an evaluation of stressors to these major attributes. For the intensive assessment (Level 3 or
L3) conducted on one-third (88) of the sites, we focused on vegetation measures and metrics. Crews
recorded all species and their cover in a 0.1 ha plot. Stem diameters and density for all live and dead
tree stems > 10 cm dbh were also collected.
All data were entered and managed in an Ecological Observations Database that was specifically
designed for the project, yet structured as generically as possible to provide an ongoing database tool
for other ecological integrity assessment projects. The database is structured to match field data
protocols: General Site Description, L 2 metrics, L 2 stressor checklists, and L3 metrics, including
vegetation plot data. Data in 2009 were stored separately for IN and Ml. In 2010, several changes were
made to the protocol, particularly for stressors checklists, necessitating a slightly different design. The
2010 data from both states were managed in a single database. An Index of Ecological Integrity (IEI),
including a scorecard, was used within the database to summarize all metric ratings for L2 assessments.
Components of the Floristic Quality Index (FQI) were used to assess integrity for L3. Data were exported
from the database in formats suitable for statistical analysis. We created a Human Stressor Index (HSI)
based on aggregating stressor scores for Hydrology, Soils, and Buffer. Our primary analyses consisted of
screening the metrics based on redundancy among metrics and discriminatory power in relation to the
HSI classes. Data are available from NatureServe and from the Natural Heritage Programs upon request.
Based on redundancy analysis of metrics, we found that two pairs of metrics had high redundancy
(connectivity vs. land use index, and native plant species cover vs. invasive plant species cover).
Conversely, among the vegetation metrics, organic matter and increasers had the lowest correlations
with other MEFs and to FQI metrics. This suggests that they were not useful metrics for assessing
ecological integrity. Based on discriminatory power, soil disturbance and water quality poorly
differentiated sites among the HSI classes. Almost all vegetation metrics had low scores in
discriminating among HSI classes, which may reflect both lack of discriminatory power to abiotic
stressors and responses to other, biotic stressors (e.g., logging, deer browse). The single vegetation
metric most responsive to the HSI was vegetation composition. Despite these individual issues with
metrics, the overall Index of Ecological Integrity (IEI), the Rank Factors (Landscape Context, Size and
Condition), and the Hydrology and Vegetation MEFs were effective in discriminating among HSI classes.
Only Soils was not.
-------
Considering the above, we dropped three metrics from our assessment (organic matter, increasers-
cover, and water quality). We kept the land use index, so we can test it across a greater variety of land
uses across the country. We suggest that native plant species cover and invasives species cover could
be treated as two parts of a Native-Invasive Species Index. We also did not drop the Soil Disturbance
metric, because we would like to test it on wider range of degraded wetlands, where greater levels of
soil disturbance may be expected. But we gave Soils only half the weight of the other two MEFs. Our
redesign provides a more equal set of metrics across all MEFs than the original design (especially for
vegetation, where the 7 metrics are reduced to 4). Our final recommended list of metrics for ecological
integrity assessments of wetland are summarized in the table below (including a tidal wetland option):
RANK FACTORS
LANDSCAPE
CONTEXT
SIZE
CONDITION
MAJOR
ECOLOGICAL
FACTORS
LANDSCAPE
BUFFER
SIZE
VEGETATION
HYDROLOGY
SOIL
METRICS
Connectivity
Land Use Index (optional)
Barriers to Landward Migration
(optional tidal)
Buffer Index
Relative Patch Size (ha) (optional)
Absolute Patch Size (ha)
Vegetation Structure
Regeneration (woody)
Native Plant Species Cover
Invasive Exotic Plant Species Cover
Vegetation Composition
Water Source
Hydroperiod
Hydrologic Connectivity
Physical Patch Types
Soil Surface Condition
Applying the Final Model
As a final check on the consistency of the method with the best professional judgment methods of
earlier Heritage methods (EORANK), we compared the IEI scores with an independent rescoring of the
same sites by Heritage staff in Michigan, who rate Condition, Size, and Landscape Context, as well as
assign an overall EORANK. We found that the Vegetation MEF of the IEI had a very high correlation with
the Michigan Condition rating. But other correlations were weaker. We found that the EORANK
methods relied more strongly on vegetation, and less on landscape context, hydrology and soils than the
IEI does. We suggest that an overall IEI is the most reliable way to evaluate current conditions of a
wetland, including both biodiversity and ecosystem processes.
In conclusion, we demonstrated that our multi-metric EIA method can be effectively used in the field to
establish a general index of ecological integrity, in a practical, ecologically meaningful way. Although
-------
some of our metrics require greater expertise than others, all attributes have at least two metrics that
can be evaluated in a relatively straightforward manner, allowing for wide applicability. The method will
have great value for the Natural Heritage Network, contributing to a consistent evaluation of reference
sites and the potential for establishing a network of reference standard (minimally disturbed) sites
within and across states. Many of these metrics are also in use by other standardized rapid assessment
methods (RAMs), including the USA RAM that is part of EPA's National Wetland Condition Assessment.
Results here can be used to both refine those methods and provide compatible information on wetland
condition across programs. And, in so far as evaluating ecological integrity is a goal within other
programs, the EIA method can be a component of those programs, including for inventory, ambient
monitoring of wetland condition, and wetland mitigation and restoration.
-------
SECTION A: THE ECOLOGICAL INTEGRITY
ASSESSMENT METHOD
A.I INTRODUCTION
Ecosystem monitoring and assessment programs are critical for resource management, given how
ecosystems vary in type, size, landscape settings, and the dramatic losses and degradation that have
occurred. These programs are increasingly addressing not just the loss of native ecosystem acres, but
also their condition. Data on the ecological condition of ecosystems can be used for ambient monitoring
of status and trends, to prioritize sites for conservation or restoration, guide mitigation applications at
site and watershed or landscape scales and contribute to land use planning (Fennessy et al. 2007, Faber-
Langendoen et al. 2008).
Agencies and organizations are in need of ecological methods that can address this aspect of
ecosystems. For example, as part of the National Wetland Condition Assessment in 2011, the
Environmental Protection Agency (EPA) carefully designed a comprehensive field survey methodology to
assess wetland condition, relying on a reference site approach to establish the criteria for wetland
condition (USEPA 2011). They were able to draw on a growing body of assessment methods that
provide standardized field sampling and reporting methods for assessing ecological condition (e.g., Mack
2001, 2004, Herrick et al. 2005, Pellant et al. 2005, Collins et al. 2006, Fennessy et al. 2007).
There are a number of ways to approach condition assessments, and it is important to clarify the
conceptual bases for doing so, in order to ensure that the methods address their intended goals. One
important basis on which to assess condition is that of ecological integrity (Andreasen et al. 2001).
Building on the related concepts of biological integrity and ecological health, ecological integrity is a
broad and useful endpoint for ecological assessment and reporting (Harwell et al. 1999). Ecological
integrity can be defined as "an assessment of the structure, composition, and function of an ecosystem
as compared to reference ecosystems operating within the bounds of natural or historic disturbance
regimes" (adapted from Lindenmayer and Franklin 2002, Young and Sanzone 2002, Parrish et al. 2003).
"Integrity" is the quality of being unimpaired, sound, or complete. To have integrity, an ecosystem
should be relatively unimpaired across a range of characteristics and spatial and temporal scales (De Leo
and Levin 1997). This broad definition can serve as a guide to developing assessment methods, steering
us through the related assessment methods for ecological functions or ecosystem services (Dudley et al.
2005, Jacobs et al. 2010).
Ecological integrity concepts are similar to the Index of Biotic Integrity (IBI) concept for aquatic systems.
The original IBI interpreted stream integrity from twelve metrics reflecting the health, reproduction,
composition and abundance of fish species (Karr and Chu 1999). Each metric was rated by comparing
measured values with values expected under relatively unimpaired (reference standard) conditions, and
the ratings were aggregated into a total score. Building upon this foundation, others suggested
interpreting the integrity of ecosystems by developing suites of indicators or metrics comprising key
biological aspects of ecosystems, such as Vegetation IBIs (Mack and Kentula 2010), or more broadly to
included, biological, physical and functional attributes of those ecosystems (Harwell et al. 1999,
Andreasen et al. 2001, Parrish et al. 2003).
To be effective, the EIA method needs to account for the wide range of ecosystem types (ultimately
including terrestrial, freshwater and marine systems), the need for various levels of assessment (remote
-------
sensing and field based, both rapid and intensive sampling methods), and the availability of a wide set of
indicators. Here we address terrestrial (dryland and wetland) systems.
Critical to this endeavor is the use of conceptual models that highlight ecological attributes for which
indicators of integrity are most needed. A conceptual ecological model delineates linkages between key
ecosystem attributes and known stressors, or agents of change. It helps identify the ecological
attributes we most need to understand regarding the ecological dynamics of the ecosystem, and which
we must address when making management decisions to maintain ecological integrity (Noon 2003).
Our goal is to present an overview of our ecological integrity methods, including a) the role of
conceptual models and indicators, relying in part on understanding their ranges of natural variability, b)
selection of indicators that assess the main ecological attributes and help inform changes that reflect
degradation, c) consider indicators at multiple levels of assessment (remote, rapid, intensive), and d)
scoring and integrating the indicators in an index of ecological integrity through a scorecard.
A.2 BACKGROUND
NATURAL HERITAGE METHODOLOGY
For over twenty-five years, NatureServe has advanced approaches for documenting the viability and
ecological integrity of individual occurrences of species and ecosystems,1 often referred to as the
"elements" of biodiversity (Stein et al. 2000, NatureServe 2002, Brown et al. 2004). Natural Heritage
methodology often uses the term "element occurrence rank" (EO rank) when referring to the ecological
integrity of these ecosystem element occurrences (EOs).2 Earlier methods relied on fairly qualitative,
expert-driven protocols. More recently, the NatureServe methodology has been revised to better reflect
an indicator-based approach, one that emphasizes specific indicators to assess the ecological integrity of
aquatic, wetland, and dryland ecosystems. Previous publications have provided some of the background
(Faber-Langendoen et al. 2008, Tierney et al. 2009, Unnasch et al. 2009); here we provide a major
overview to the methods.
PROJECT ORGANIZATION
The project was funded by the Environmental Protection Agency (EPA). The primary organizations
involved in the project are NatureServe, the Michigan Natural Features Inventory (MNFI) and the
Indiana Natural Heritage Program (INNHP), with data being made available to Michigan Department of
Environmental Quality and others. The development of the Ecological Integrity Assessment (EIA)
method, including the conceptual model, occurred over a number of earlier projects, but we provide
important updates here, including its application across all three levels of assessment (remote, rapid,
Natural Heritage methodology was originally developed by "Natural Heritage" staff of The Nature
Conservancy (TNC), many of whom then transferred to NatureServe when it was formed in 2000. Since then,
NatureServe staff have worked with the Network of Natural Heritage Programs to maintain and improve the
methodology, while continuing to collaborate with TNC staff.
2 In addition to calling ecological integrity an "element occurrence rank" (or EORANK), Heritage methodology also
refers to ecological integrity criteria as "Element Occurrence Ranking Specifications" or EORANKSPECS. Occurrence
requirements and mapping guidelines are referred to as "Element Occurrence Specifications" or EOSPECS. We
introduced the term "ecological integrity assessment" because it is the more widely used term in conservation
biology.
-------
and intensive). This project was coordinated by NatureServe, whose staff served as the Principal
Investigators, with field work contracted to the MNFI and INNHP staff. In addition, staff from EPA
provided project oversight, regional input, planning, and technical assistance. Staff from EPA's National
Wetland Condition Assessment team and from the Michigan Department of Environmental Quality
provided feedback on various aspects of the project.
A.3 PURPOSES OF ECOLOGICAL INTEGRITY ASSESSMENTS
The goal of an ecological integrity assessment is to provide a succinct assessment of the current status
of the composition, structure and processes of a particular occurrence of an ecosystem type. These
assessments may be done for a number of purposes, including:
• Prioritize occurrences for conservation/management actions (often filtered through site selection
criteria).3 Ratings are helpful both as absolute ratings (best anywhere) and as relative ratings (best
of what we have).
• Track status of occurrences over time. After a site is protected and/or put under management,
there is a need to know whether the integrity of the occurrence is staying the same or changing.
Cost-effective, reliable measures of integrity are needed (Tierney et al. 2009).
• Contribute to information on conservation status. The condition or integrity of occurrences are
considered when assigning global, national, and subnational4 conservation status ranks (GRANKs,
NRANKs, and SRANKs), because knowing how many good quality occurrences are on the landscape
is an important guide to the overall conservation status or at-risk status of an ecosystem.
Conservation status could also be assessed at other scales.
• Prioritize field survey work. Ratings may be used effectively to guide which occurrences should be
recorded and mapped (see NatureServe 2002, Section 6, EO Tracking), and to help prioritize
occurrences for purposes of conservation planning or action, both locally and rangewide.
• Assess restoration/mitigation efforts. There is an increasing value in using benchmark sites
(reference sites) of known integrity to set performance standards for restoration and mitigation - to
ensure that wetlands are restored to desired conditions (Faber-Langendoen et al. 2008).
• Inform species population viability ranks. When species populations are closely linked to specific
wildlife habitats or ecosystem types, the occurrence rank of the habitat type may serve as a guide
for the species viability ratings.
3 Although Element and Element occurrence (EO) ranks help to set conservation priorities, they are not the sole
determining factors. The determination of priority occurrences for conservation action will include not only the
conservation status of the Element and the likelihood of persistence of the occurrence, but will also include
consideration of other factors such as the taxonomic distinctness of the Element; the genetic distinctness of the
EO; the co-occurrence of the Element with other Elements of conservation concern at a site; the likelihood that
conservation action will be successful; and economic, political, and logistical considerations.
4ln this document, the term "subnation" will refer to the first order subdivision of a nation (e.g., state, province,
district, department).
-------
A.4 THE UNIT OF ASSESSMENT- ECOSYSTEM OBSERVATIONS
E COSYSTEM TYPES
The term "ecological integrity assessments" can refer broadly to many aspects of the natural world.
Including species (here perhaps the term population viability is more widely used) and other focal
resources (Unnasch et al. 2009). Our focus here is on ecosystem types, at multiple scales. On the one
hand, we use the term "ecosystem" in a general sense to cover a wide range of scales from macro-scale
(such as formations and biomes or broad wetland classes), through mesoscale units (such as
macrogroup, group, ecological systems, order, regional wetland classes), to micro-scale units (alliances,
associations, natural community types etc) (see Faber-Langendoen et al. 2008). On the other hand, we
use ecosystem in a particular, and pragmatic sense, as a spatial entity bounded on the landscape by a
specific set of biotic and abiotic structural, compositional and ecological criteria, at scales that address
the most common concerns of conservation and management.
We refer to the specific place where an ecosystem type is found as an "ecological observation." Many
other terms are used, such as "assessment area," "sample point", "ecological site," "field site,"
"occurrence," or "stand." The term "observation" is sometimes used as a generic, flexible term applied
to any kind of place or unit where an ecosystem is identified and described (Stevens and Jensen 2007),
and is increasingly used a term for all species or ecosystem field records (Lapp et al. 2011). The EIA
method focuses on understanding the structure, composition, and processes that govern the wide
variety of ecosystem types at particular sites where an ecosystem is found. Ecosystem classifications are
important tools in assessing the ecological integrity of these observations. They help ecologists to
better cope with natural variability within and among types so that differences between observations
with good integrity and poor integrity can be more clearly recognized. Classifications are also important
in establishing "ecological equivalency," for example, in providing guidance on how an impacted salt
marsh can be restored to a salt marsh with improved integrity.
POINTS, POLYGONS, AND PATCHES
Assessments of ecosystem condition can be based on observations defined as points, polygons, or
patches.
A point based approach in which a fixed area is sampled around a point offers some advantages
(Fennessy et al. 2007, Stevens and Jensen 2007):
• simplicity in terms of sampling design
• no mapped boundary of ecosystem type is required for assessment unit
• limits practical difficulties in the field of assessing the entire area, as the area is typically
relatively small (0.5-2 ha). Long term ambient monitoring programs often use a point-based
approach because of these advantages.
A polygon approach, in which a specific ecosystem area is delineated (using vector or raster methods),
offers some advantages:
• Mapping boundaries facilitate whole ecosystem and landscape interpretations
• Decision makers and managers are often more interested in "stands" or "occurrences," rather
than points.
• Programs that maintain mapped occurrences of ecosystem types are most interested in the
status and trends of those occurrences.
10
-------
Pixel (or raster) based imagery, such as from satellites, are perhaps intermediate between points and
polygons. Pixels are often smoothed into larger "patches," these patches can be assigned to ecosystem
types, and analyses can be performed on these patches, or on a series of patches. These series may be
created as clusters (e.g., using separation distances between patches, comparable to cluster polygons)
or as "bounded patches," where a larger landscape or watershed boundary is used, and all patches of
the same ecosystem type within that boundary are included as part of the assessment area. The
"bounded patch" approach is currently being used by NatureServe to conduct ecological integrity
assessments in western U.S. ecoregions (NatureServe 2011 in prep).
A raster or vector based polygon approach is widely used for inventory purposes by many agencies.
Many programs that are part of Natural Heritage Network routinely classify and map ecosystem
occurrences, and maintain extensive information on the structure, composition and stressors to those
occurrences. While not without its complications, standard mapping guidelines can be applied and line
work can be provided digitally to crews, who can then adjust them as needed directly on aerial photo or
digitally on field recorders.
There are also regulatory advantages to polygons, as noted by Fennessy et al. (2007) for wetlands "the
basic "currency" in Clean Water Act Section 401/404 regulation of wetlands is something called a
"wetland" and this is also the common understanding: a "wetland" is a definable piece of real estate
that can be mapped and walked around. There are substantial pragmatic and legal considerations in
developing a condition assessment protocol that cannot assess "wetlands."
Where the occurrence at a site is the focus, then a sampling design could still vary as follows:
• conduct an assessment survey of the entire area of the occurrence, e.g., a rapid qualitative
assessment;
• conduct an assessment survey of a typical sub-area(s) of the occurrence, preferably of uniform
condition, or
• collect a series of plots, placed either in representative or un-biased locations, throughout the
entire area or sub-area occurrence.
In all three cases, the intent is to assess the ecological integrity of a particular wetland occurrence. The
occurrence may, in fact, be defined by the combination of ecosystem type and level of integrity. Thus a
minimally-disturbed wetland type can be mapped and assessed separately from a degraded example.
WATERSHEDS & LANDSCAPES
The condition of entire watersheds or ecoregions can also be assessed. Ecoregional status assessments
or watershed profiles are ways of characterizing the entire landscape area. We do not apply the term
"ecological integrity assessment" to these approaches, as our definition of ecological integrity is "an
assessment of the structure, composition, and function of an ecosystem as compared to reference or
benchmark ecosystems operating within the bounds of natural or historic disturbance regimes."
Although ecoregional units can be viewed as "landscape ecosystems (Bailey 1996)," these units more
often are viewed as landscape or watershed types within and across which ecosystems are found
(Forman 1995); i.e., the ecosystems provide a "bottoms-up" approach where assessments of component
ecosystems contribute to an overall rating for the landscape or watershed. Nonetheless, assessing the
condition of the landscape area can provide important information on the ecological integrity of the
ecosystems within those landscapes.
11
-------
A.5 RANGE OF NATURAL VARIABILITY AND ECOLOGICAL INTEGRITY
THE RANGE OF NATURAL VARIABILITY CONCEPT
Species and ecosystems all evolve within dynamic environments; and naturally exhibit some range of
natural variability in their attributes over time and space. For example, the age and species composition
of any forest canopy naturally vary over time and from one stand to the next; and any forest naturally
experiences varying frequencies and intensities of disturbance from fire, drought, wind damage, or
flooding. Similarly, coastal salt marshes naturally experience varying frequencies and intensities of
nutrient and sediment inputs, tides, wave action, and storms. Within the limits of this range, further, the
variation may be either patterned (e.g., cyclical) or random; and may play out over scales of time from
hours and days to decades and centuries.
The natural variation in both space and time is thus essential to shaping ecosystems. Consequently, the
natural range of variability depends on specifying the time frame. For purposes of assessment projects,
where the horizon is usually 30 to 100 years, we normally treat the natural variability in each key
attribute of a system as occurring within stable limits. However, there may be situations in which this is
not appropriate.
Resource managers often use the concept of a range of natural variability (RNV) (e.g. Landres et al.
1999, Oliver et al. 2007). Information on RNV provides important clues on the long term driving
variables and disturbances that shape ecosystems, the flux and succession of species, and the relative
role of humans in shaping the systems. Understanding RNV is important for placing interpretations of
ecological patterns in their historical setting. However, what is 'natural' can be difficult to define, given
limited knowledge of ecosystems, the extent of past human activity, and the likely effects of ongoing
and future climate change. Scientific knowledge of most ecosystems has a relatively short history, as
does the preserved record of most environmental regimes (fires, floods, etc.). The variation in ecological
dynamics that we observe within years or decades can be part of much larger trends or cycles spanning
centuries or millennia. For these reasons, others prefer the term "historic range of variability" (Egan and
Howell 2005).
No ecosystem, natural community, or species is ever static when viewed on larger scales of time. Human
activity has thoroughly transformed many places throughout the world, and no place is free of human
impacts. Much as a changing climate throughout the Holocene (past 12,000 years) brought about
changes in many of aspects of ecosystems, and resulted in many patterns of species composition we see
today, so too have certain human activities shaped ecosystems. Humans have brought about large-scale
and long-term changes in ecosystems even far from our farms and cities, for example through hunting
and selective tree removal, releasing non-native species, setting fires, and diverting streams.
In many instances where the rate and magnitude of human-induced change may be limited, we can
safely subsume their effects within a practical 'natural' range of variation. That is, we can assume that
their effects have had only a limited impact on the evolutionary environment of biodiversity. At the
same time, we can often detect human effects that cause rapid and substantial ecological change. And
we can do so not only in recent, better documented times but in the more distant past, for example
from records of ancient land clearing for corn production, desert stream diversions, or the draining of
arable swamplands. When we can detect such more significant human effects, we need to presume
them to be outside of some practical, ecologically functional range of natural variability (i.e., likely
resulting in local extinctions and other biodiversity impacts).
12
-------
ECOLOGICAL INTEGRITY AND RANGE OF NATURAL VARIABILITY
Given these challenges, it is important to emphasize what can and cannot be achieved by using RNV as a
component of ecological integrity (Higgs and Hobbs 2010). First, it is the knowledge of natural variability
that informs our goals and evaluations of current conditions, but this knowledge does not a priori
constrain how we state desired conditions (clarity in goals). Second, to suggest that we can simply take
over the management of natural ecosystems without understanding RNV is to invite failures in these
complex systems (restraint and respect). Finally, the purpose of understanding RSV is not to lock us in
the past, but to ensure that we connect the historical ecological patterns and processes to the present
and future (historical fidelity). Fourth, understanding RNV will ensure that we can anticipate change and
emphasize resilience in the face of future changes (Higgs 2003, Higgs and Hobbs 2010). In this way, NRV
as a component of ecological integrity takes us beyond a simple interpretation of what is natural to
engaging us to think through how our actions and goals can maintain natural ecosystems. Finally, NRV
need not be interpreted solely from the historical record, but from benchmark sites currently present on
the landscape.
Because it can be difficult to define what is natural, alternative terms have been suggested, including
"acceptable range of variation" (Parrish et al. 2003). The key point is that direct knowledge of the range
of natural variability, is but one source of information for developing proposed ratings of ecological
integrity. Other sources of information include ecological models, expert knowledge, and comparisons
to a reference gradient or reference standard of the same or similar ecosystems (Parrish et al. 2003,
Stoddard et al. 2006). Even where present-day reference standard sites may be hard to identify based
on minimally disturbed criteria, one may still be able to make reasonable estimates based on historic
data or inferred species-habitat relationships (Brewer and Menzel 2009). Particularly where such
examples have been affected by human impacts of varying types and magnitudes, comparisons can be
especially informative about where the limits may lie beyond which the persistence of the ecosystem
may be at risk. Even where present-day reference standard sites may be hard to identify based on
minimally disturbed criteria, one may still be able to make reasonable estimates based on historic data
or inferred species-habitat relationships (Brewer and Menzel 2009).Thus the ranges of ranges of
variability specified in the indicators are ranges relevant to the hypothesized levels of ecological
integrity, and our understanding of those ranges will change over time.
Thus, both our understanding of ecological integrity and what is natural change over time. Too often
the characterization of integrity is treated as a static linear function, not unlike the model shown in Fig.
Al. In the short term, these models can be helpful. But they can be mis-leading with respect to both
the ongoing natural, historical processes that shape ecosystems and the human interactions with those
systems. Simplistic views of "natural" as referring only to "pristine conditions" is not tenable, given the
long interactions between humans and the environment. But simply collapsing human activity (culture)
into an extension of natural processes is also too simplistic. It may be helpful to expand our view by
considering how ecology and human culture are "knitted together over time;" that is, both culture and
ecology have histories, and consideration of current ecological integrity reflects both histories, without
suggesting that they are one and the same (Fig. A2, Higgs 2003). What is critical is to ground our ideas
of ecological integrity in the knowledge gained from current reference sites; thereby spanning our
cultural perspective on integrity with known ecosystem sites in the present, as informed by the past.
13
-------
FIGURE Al. Simple schematic showing how ecosystem structure and function may recover over
time to either the more original (historical, natural) system or some altered form.
LLJ
Original Ecosystem
or Goal
Degraded or Present
State
Species & Complexity
Ecosystem
Structure
FIGURE A2. A model of ecosystem change showing how ecosystems and culture are inter-related
through time. Red/orange color highlights the increasing changes affecting ecosystems, and the
uncertainty of those changes into the future (Adapted from Higgs 2003, Fig. 6.2).
Ecosystem
future
O
Assessment methods
Ecological History
Cultural desired
conditions
Present
Ecological Integrity ^V Reference conditions
- pristne to
degraded
Cultural History
Human Culture Ecosystems
ECOLOGICAL INTEGRITY AND CLIMATE CHANGE
Global climate change, a further consequence of human activity, is bringing about changes in regional
and local climate. Every place on Earth now faces changes in the magnitude, timing, frequency, and
duration of atmosphere-driven conditions - from changes in seasonal temperatures and weather
patterns to changes in the temperature and pH of our oceans - many potentially outside the range of
14
-------
historic variation. The ecosystems of tomorrow in every region potentially will experience ranges of
variation in atmosphere-driven conditions far different than have the ecosystems in these regions even
in the recent past.
Understanding the natural or historic range of variability has immense value in improving our
understanding of ecosystem responses to environmental changes and setting management goals (e.g.,
Swetnam et al. 1999). However, as it now apparent from our discussion above regarding ecological
integrity, it may no longer be sufficient to assume that establishment of historical conditions will assure
that ecosystems will persist into the future, in the face of novel anthropogenic stressors, such as
pollution, habitat fragmentation, land-use changes, invasive species, altered natural disturbance
regimes and climate change (Millar et al. 2007). Thus measures of ecological integrity need to account
for the ability of ecosystems to "adapt" to changes, as climate and environments shift. These shifts may
create environments that are outside any known range of natural variation ("novel climates" of Williams
et al. 2007). This does not make the past and current states irrelevant; rather, as Millar et al. (2007)
note: "Historical ecology becomes ever more important for informing us about environmental dynamics
and ecosystem response to change."
A.6 ECOLOGICAL INTEGRITY ASSESSMENT METHOD
Our method is based on the following set of key steps:
1. determine the purpose of the assessment
2. develop a general conceptual model for wetlands, adapted, as needed, for various ecosystem types
3. rely on indicators of ecological attributes that span the major structural, compositional and
ecological processes of the system
4. select indicators across three levels of assessment - (i) remote sensing, (ii) rapid ground-based, and
(iii) intensive ground-based metrics
5. scale the thresholds or assessment points of the indicators based, in part on ranges of natural and
historic variability, ecological models, benchmark or reference sites
6. summarize indicators using ratings and integrate into an overall index of ecological integrity
PURPOSE OF THE ASSESSMENT
In the section above, we noted some broad purposes of ecological integrity assessments, such as, 1)
prioritizing observations for conservation/management actions, 2) tracking status of observations over
time, 3) assess management actions, 4) contribute to range-wide conservation status of ecosystems, and
5) provide performance standards for mitigation and restoration. These general goals need to be
further refined to make sure the assessment is structured to address the needs. The geographic scale of
the assessment also has an impact e.g., national trends monitoring, regional landscape assessment, local
landscape assessment, local site management and monitoring. These varied purposes require that the
approach to ElAs be flexible, while retaining a consistent core of ecological attributes. To do so, we
develop a general conceptual model, and then suggest ways in which its application can be applied in a
flexible manner.
CONCEPTUAL MODEL FOR TERRESTRIAL ECOSYSTEMS
Identifying the ecological attributes that need to be assessed involves building a conceptual ecological
model of ecological integrity. This model rests on the knowledge of the system, its setting, and similar or
15
-------
associated systems. The result is a set of hypotheses about how the system functions, its defining
characteristics and dynamics, and critical environmental conditions and disturbance regimes that may
act as drivers of these characteristics and dynamics. These hypotheses both guide management and
monitoring, and highlight gaps in knowledge that require additional investigations (Unnasch et al. 2009).
We use a conceptual ecological model that provides a general set of ecological factors common to all
terrestrial (wetland and upland) systems, and then encourage identification of individual key ecological
attributes for individual system types. The model also provides a means to assess stressors or agents of
change to the ecological factors (Noon 2003). The terms for the model come from a variety of models
available in the literature (Table Al) and that of Faber-Langendoen et al. (2006).
TABLE Al. Comparison of terminology among various agencies and organizations for ecological integrity
/ assessments (modified from Faber-Langendoen et al. 2006). Overarching goals and objectives are
defined variously by each group. TNC = The Nature Conservancy, EPA = Environmental Protection
Agency, NPS = National Park Service.
NatureServe
Rank Factor
Major Ecological
Factor
Key Ecological
Attribute
Indicator/Metric
Comer et al.
(2003), Faber-
Langendoen et al.
2008)
TNC
Key Ecological
Attribute
Indicator
Parrishetal. 2003
EPA
GoalObjective
Essential Ecological
Attribute (EEA)
EEA subcategory
Ecological Endpoint
Measure
Harwell et al 1999,
Young and Sanzone
2002
NPS Vital Signs
Level 1 Category
Level 2 Category
Level 3 Category
(Vital Sign)
Measure/ Metric
Fancy etal. 2009
The major components of the model include three primary rank factors (landscape context, size, and
(on-site) condition, subdivided into 6 major ecological factors: landscape, buffer, size, vegetation,
hydrology, and soils. Together these are the components that capture the structure, composition and
processes of a system (Fig. A3). Other major attributes, such as birds, amphibians, and
macroinvertebrates, can also be assessed where resources, time and field sampling design permit. The
model is fairly intuitive, but a key component is that, to describe how a system "works," one must
include both the "inner workings" (condition) and the "outer workings" (landscape context). Assessing
size of ecosystems helps to characterize patterns of diversity, area-dependent species, and resistance to
stressors. Conservation of such characteristics and processes will contribute not only to current
ecological integrity but to the resilience of the ecosystem in the face of climate change and other global
causes of stress.
16
-------
FIGURE A3. Example of conceptual model for ecological integrity assessments of terrestrial
ecosystems. The core ecological attributes of ecological integrity are shown for wetland and uplands.
The model can be expanded to include additional measures of biotic integrity, such as birds,
amphibians, macroinvertebrates, etc.
Rank Factors and Major Ecological Factors
An ecological factor (sometimes referred to as a major ecological attribute (Faber-Langendoen et al.
(2008) of an ecosystem is common to all ecosystems and encompasses a broad set of related ecological
attributes. Thus it serves to ensure that assessment methods will not neglect major structural,
compositional and ecological processes of the system. These ecological factors can also be used as
broad indicators in their own right, and may be assessed qualitatively, using narrative approaches, in
very rapid assessments.
For our wetlands model, we use three primary rank factors- Landscape Context, Size, and Condition, as
is typical of a number of major assessments (NatureServe 2002, Parkes et al. 2003, Oliver et al. 2007)
Together these primary rank factors provide the orientation for more specific selection of major
ecological factors, key ecological attributes and indicators
The primary and major ecological rank factors are:
• "Landscape Context" refers both to the spatial structure (spatial patterning and connectivity) of
the surrounding landscape and the immediate buffer within which the ecosystem occurs; and to
critical processes and environmental features operating at landscape scales that affect the
system. Examples of landscape structure include attributes of fragmentation, patchiness, and
proximity or connectivity among habitats. Examples of landscape processes include the
17
-------
interaction of matter, energy, and disturbance regimes between a focal ecological system and
surrounding systems.
"Size" refers to attributes related to the absolute or relative size of the ecosystems, measured
as geographic extent, or area, or the size of the natural patch within which a system is found.
"Condition" refers to critical biotic and abiotic structure, composition, and processes, and their
dynamics, typically subdivided into vegetation, hydrology and soils. Examples include species
composition, hydroperiod, soil chemistry, grazing intensities, and vegetation structural stages
arising from natural disturbances.
Although the model presumes that all systems share the primary and major ecological factors, the
weights given to them may vary. E.g., hydrology can be given much more weight in wetland systems.
Size may be of minor importance in spatially patchy wetland systems. Many standard wetland
assessments emphasize these ecological factors - buffer, vegetation, hydrology, and soils (Fennessy et
al. 2007). Consistent recognition of these ecological factors also helps interpret the information coming
from a variety of assessment programs, particularly as these vary in the role given to interpreting the
biotic versus abiotic assessment of condition. Programs that use Indices of Biotic Integrity are relying on
the biota or vegetation as the primary or sole measure of condition. Natural Heritage Programs often
use vegetation as the primary means of assessing condition. For landscape context, many programs
measure both the buffer around a point or polygon and the larger landscape (e.g., Jacobs et al. 2010).
Size provides some challenges to an integrity assessment. Some ecosystems types vary widely in size for
entirely natural reasons (e.g., a forest type may have very large occurrences on rolling landscapes, and
be restricted to small occurrences on north slopes or ravines in other landscapes). For other types, size
may be relatively unimportant because they are always small (seepage fens, talus slopes). Nonetheless
size can be an important aspect of integrity. For some types, diversity of animals or plants may be
higher in larger occurrences than in small occurrences that are otherwise similar. For occurrences that
occur in mosaics, the larger occurrences often have more habitats. Larger wetlands are also more
resilient to hydrologic stressors, since they buffer their own interior portions to some extent. Studies
have also shown that wetland size is a strong predictor of wetland condition, probably as a function of
landscape fragmentation (Fennessy et al. 2009). Thus size can serve as a readily measured proxy for the
interdependent assemblage of plants and animals, particularly in the context of more rapid
assessments, where limited measurements can be taken. Finally, we can separately assess absolute size
from relative size (the size a patch would have under natural conditions versus stressed condition (e.g., a
small wetland that has been partially filled in or drained). For all these reasons, we retain size as a high
level indicator, but keep it separate from other indicators, so that its contribution in the overall rating of
integrity is clear.
Recognized stressors (threats) to an ecosystem also provide crucial information for the identification of
key ecological attributes. Stressors to the system include human activities, structures, or institutions -
or consequences of these. They alter one or several key ecological attributes beyond their acceptable
ranges of variation. Consequently, knowledge of how specific human actions cause harm to an
ecosystem can provide insight into the resource's key ecological attributes, and vice versa. Again, to
simplify these steps and to ensure consistency across models, we suggest that stressors should be
tracked across all ecological factors, but can be fine-tuned based on knowledge of key ecological
attributes.
18
-------
Key Ecological Attributes and Indicators
A key ecological attribute (KEA) of an ecosystem is one that is critical to a particular aspect of the
ecosystems persistence in the face of both natural and human-caused disturbance, and alterations to
that attribute beyond some critical range of variation will lead to the degradation or loss of that
ecosystem. It is much more directly measurable than are ecological factors and more amenable to an
indicator-based approach.
Metrics are sometime referred to as indicators, and here we use the term in the sense of a fine-grained
indicator. For clarity, we distinguish "metrics" from both "measures" and "general indicators."
Measures are those values that are collected directly in the field (e.g., diameter of tree at breast height,
species percent cover) and metrics are values derived from specific measures (e.g., basal area, stand
structural class, species diversity) that inform us about the status of an ecological attribute of integrity.
Coarse woody debris is a general indicator, whereas volume of coarse woody debris and biomass of
coarse woody debris are two closely related metrics for that indicator. For a given system, a single
metric is typically used for an indicator, but when comparing across systems, we may have different
metrics for the same indicator. Standardization of metrics is certainly helpful, but more important are
agreement on the kinds of general indicators that are used to characterize ecological integrity, at least
among related ecosystem types.
To be a metric (or specific indicator) requires that it be informative about alterations to the attribute
that may lead to the degradation or loss of the ecosystem.
The metric may be either:
• A specific, measurable characteristic of the major or key ecological attribute (e.g., percent cover
of native species, coarse woody debris, calcium:aluminum ratio of soils, hydroperiod).
• A collection of such characteristics combined into a "multi-metric" index, such as a forest
structural stage index that integrates measures of tree size across all stems, or a buffer index
that combines buffer length, width and condition, or
• A measurable effect of the key ecological attribute, such as a ratio of the frequencies of two
common taxa of aquatic insects (the indicator) that varies with changes in average Nitrate
concentration (the key attribute) in a stream.
Metrics are selected to meet ecological, technical and management needs (Tierney et al. 2009, Unnasch
etal. 2009, Fancy 2009).
Ecological - Statistic Criteria
1. Specific (redundancy): unambiguously associated with the key or ecological factor of concern
and not significantly affected by other factors.
2 . Sensitive (discriminatory power): able to detect changes that matter to the persistence of
the ecosystem.
3. Comprehensive (range): able to detect changes across the entire potential range of variation
in the key ecological attribute, from best to worst condition.
Technical Criteria
19
-------
4. Measurable: measurable by some procedure that produces reliable, repeatable, accurate
information.
5. Technically feasible: amenable to implementation with existing technologies without great
conceptual or technological innovation.
Management Criteria
6. Timely: able to detect change in the key ecological attribute quickly enough that project
managers can make timely decisions on conservation actions.
7. Cost-effective: able to provide more or better information per unit cost than the alternatives.
8. Partner-based: compatible with the practices of key partner institutions in the conservation
effort, or based on measurements they can or already do collect.
9. Legal mandates: addresses legal requirements for the program.
It is rarely possible to identify a single indicator that meets all eight criteria for an individual key
ecological attribute, but ensuring relevance across the broader categories is important. Managers may
need to put several indicators together to obtain a more reliable or more complete picture of the
system. For example, indicators taken from both field surveys and analyses of aerial photographs may
provide complementary and more reliable assessments of forest tree composition, or of characteristics
of the buffer around a wetland, than either indicator can on its own.
A variety of statistical methods are available to help assess the statistical rigor of metrics, applicable to
both rapid and intensive metrics. The most readily assessed criteria include comprehensive range,
discriminatory power or responsiveness, and redundancy (Blocksum et al. 2002, Klemm et al. 2003,
Jacobs etal. 2010).
As an example of a good indicator is that of, "Relative Total Cover of Native Plant Species." It is
measured by estimating total cover of all exotic species subtracted from total cover of all vegetation and
divided by 100" (Table A2). It is specific, measurable, fairly sensitive (though separating invasive exotics
from other exotics may increase its sensitivity), timely (again distinguishing invasives from other exotics
might increase its timeliness), technically feasible (the "exotics" category is generally well defined in
botanical manuals), cost-effective and partner-based (many organizations and agencies are interested in
the presence and abundance of exotics). It may be one of several specific indicators for a Vegetation
Composition KEA. Similarly, "Woody Regeneration" (such as tree seedling and sapling regeneration in
forests) is another specific indicator (metric) for Vegetation Structure (KEA). Together these are
indicators for overall Vegetation (MEF). Measurement of indicators is described in protocols that ensure
consistent and repeatable measurements. By contrast, leaf and other small woody cover in systems is
much harder to specify in terms of natural variability relevant to ecological integrity, particularly in rapid
assessments. It may be too variability within a site over the course of a season.
The identification of key ecological attributes and specific indicators for each ecosystem or group of
related ecosystems is an iterative process. It may require that KEAs and indicators be identified for each
ecosystem type, or they may be applicable across many types. A review of the level of ecological
classification may be needed to ensure that KEAs and indicators are ecologically meaningful, applicable
to issues of resource management and cost-effective.
20
-------
TABLE A2. Example of a Vegetation indicator (metric)
RANK FACTOR: CONDITION
MAJOR ECOLOGICAL FACTOR VEGETATION
KEY ECOLOGICAL ATTRIBUTE Vegetation Composition
Metric: Relative Total Cover of Native Plant Species
Percent cover of the plant species that are
Definition:
native, relative to total cover (sum by species)
Metric Ratings
A = Excellent
B = Good
C=Fair
D = Poor
E = Very Poor
Metric Criteria
>99% cover of native plant species
95- 99% cover of native plant species
80-94% cover of native plant species
50-79% cover of native plant species
<50% cover of native plant species
To assist with this process, and ensure a level of standardization across models, we suggest that any
model should include key ecological attributes and indicators under each of the primary factors. The
final list should focus attention on those potential key ecological attributes that are the most defining,
most critical or pivotal to the persistence of the ecosystem and its natural internal dynamics. If an
attribute or indicator does not appear to be responsive to stressor changes or appears to be unrelated
to these major attributes, it is a signal that it is not critical to ensuring the persistence of the ecosystem.
Integrity Metrics and Stressor Metrics
The primary emphasis of the indicators is on measuring a relevant attribute of the ecosystem itself that
is clearly related to known ranges of natural variability and that are responding to stressors. We refer to
these as "integrity metrics" or indicators. We can also measure the stressors themselves, but
information from these metrics provides only an indirect measure of the status of the system - we will
need to infer that changes in the stressor correspond to changes in the integrity of the system. We refer
to these as "stressor metrics." We provide a catalogue of possible stressors at a site (stressor checklists)
to guide interpretation and possible correlations between ecological integrity and stressors.
We prefer to use integrity metrics separate from stressors, in order to independently assess the effects
of stressors on integrity, but occasionally a stressor metric is substituted for an integrity metric when
measuring integrity is challenging or not cost-effective. For example, a "Land Use Index" indicator is a
stressor metric that characterizes the level of stress produced by land uses in the surrounding
landscape, rather than characterizing the integrity of the ecosystems in the surrounding landscape. The
basic goal is an accurate, cost effective estimate of integrity, rather than concern to keep the model
pure.
Thresholds in Ranges of Ecological Integrity
The EIA method posits that some degree of thresholds exist in the range of potential variability for each
key ecological attribute. These are thresholds, outside of which managers should anticipate - or
sometimes may already observe - signs of unacceptable change or degradation to the ecosystem
(Mitchell et al. 2011). Ecologists typically cannot estimate specific probabilities of persistence for
communities and ecological systems, as can be done for species populations. Instead, we recognize that
21
-------
unacceptable alteration will involve severe degradation of an attribute or the entire system, leading to
its transformation into some other kind of system altogether (e.g., the stream flow stops, leaving a dry
stream bed; a grassland becomes a woodland in the absence of fire). Such a transformation might begin
with the loss of only a few highly sensitive species, although it could increasingly affect the more
common and less specialized as well.
The EIA method requires identifying the critical ranges of variability for each indicator used to keep track
of each key attribute, for each resource. Of most concern are stressors that cause certain ecological
attributes to vary beyond certain threshold values. Critical ("hard") thresholds occur when a given point
along a continuum of change in an ecological attributes leads to an alternative state of an ecosystem
(sensu Moiling 1973). Hard thresholds mark specific conditions beyond which ecosystems change
irreversibly; soft thresholds track a range of conditions over which ecosystems are changing, with
varying degrees of reversibility. An example of hard thresholds is when certain levels of phosphorus
loading in streams may cause a series of cascading effects. With soft thresholds, an alternative state
may emerge gradually ("soft" thresholds) as the attribute changes (Mitchell et al. 2011). For example,
declines in species richness in response to degradation may show a continuous decline, and setting a
threshold is more akin to setting benchmarks along a continuum.
When a key ecological attribute crosses either a hard or soft critical threshold, the resource itself may
not experience either immediate or abrupt change. The resource may initially only lose its capacity to
resist change triggered by new disturbances and/or its capacity to recover following a new disturbance.
Once a resource suffers such a loss of resistance or resilience, however, it may take only a slight
additional change to trigger further alteration away from its acceptable range of variation. For example,
the suppression of fire in an aspen woodland for more than a few decades could leave it vulnerable to
the arrival of seeds from other nearby communities, that could lead to the replacement of the dominant
tree cover by Douglas fir and other conifers that promote changes in soils and ground-cover vegetation
that attract different fauna that further transform community dynamics, and so forth (Unnasch et al.
2009).
The changes that ensue when a key ecological attribute passes some critical threshold may take
considerable time to play out, particularly in systems with very long-lived species. Nevertheless, once set
in motion, such chains of consequences may be difficult to reverse. The alteration of one or more key
ecological attributes beyond their acceptable ranges of variation can reach a further threshold, beyond
which the focal resource will almost certainly fail unless the situation is quickly reversed. Particularly
worrisome are thresholds of ecological collapse, failure which could mean potentially irreversible
transformation into - or replacement by - some other kind of community or system (the "thresholds of
imminent loss" of Unnasch et al. 2009).
The ease of crossing threshold may be different from a degrading perspective than aggrading
perspective. For example, having crossed the C/D threshold for phosphorus loading, it may be very
difficult to restore back to a C. Or having lost a top predator or grazer from a system, it may be difficult
to re-establish it. Or once the density of shrubs in a pine savanna increases to a certain level, it may
become very hard to reintroduce fires.
Estimating the range of natural variability and assessing thresholds for each indicator answers the
crucial questions, how much alteration of a key ecological attribute is too much? Managing ecosystems
based on NRV in turn does not mean managing for all the variation that the resource might experience
under undisturbed conditions. Instead, it means managing only for an envelope of conditions that
22
-------
together are "sufficient" for resource persistence, function, and for achieving related management
goals.
Estimating the range of natural variability for every indicator may be a challenge. It requires some
knowledge of the historic range of variability for all key ecological attributes and their indicators. Or it
could mean establishing range of variability from a current set of reference sites judged to be in various
states of ecological integrity (these can in turn be linked to historic interpretations of NRV at those sites.
Fortunately, even initial approximations about the acceptable range of natural variability for an indicator
provide hypotheses on which both to begin management and to begin research to improve the initial
estimates.
Reference Condition as a Guide to Indicator Rating
A key method for establishing ecological integrity ratings is based on reference sites. We can refer to the
full range of reference sites as the reference gradient (also referred to as reference set, or reference
network); that is "the gradient of ecosystem condition across a region varying from least disturbed
(reference standard) to highly impaired. The set of reference sites represent the range of variability that
occurs among stands of a wetland type as a result of both natural processes (e.g., succession, channel
migration, fire, erosion, and sedimentation) and anthropogenic alteration (e.g., grazing, timber harvest,
and clearing) (Klimas et al. 2006).5 There has been much discussion on whether reference standard
should be based on "minimally disturbed wetlands," i.e., the subset of the gradient of reference
wetlands that exhibit metric ratings for the type at a level that is characteristic of the historically and/or
currently minimally disturbed wetland sites in the landscapes (Klimas et al. 2006, Stoddard et al. 2006).
Using the "minimally disturbed" approach, the reference standard sites would typically have, or be able
to attain a high ecological integrity rating for all or most metrics.6 The geographic area from which
reference wetlands are selected is sometimes referred to as the reference domain (Smith et al. 1995).
The reference domain may include all (ideally), or part (e.g. within an ecoregion), of the full range extent
of a type. Where few sites exist today that are minimally disturbed, reasonable estimates can still be
made based on historic data or inferred species-habitat relationships (Brewer and Menzel 2009).
5 As Sutula et al. (2006) state, "one important element of metric development is definition of the
standard of comparison that defines the highest and lowest levels of potential or expected wetland
condition. This standard of comparison is commonly referred to as a reference; however, the concept of
reference is more accurately defined as a range of conditions that can be correlated with a known set of
stressors. The highest point on this reference continuum is then termed reference standard condition.
The collection of sites or theoretical states that represent a gradient in conditions is referred to as the
reference network. To the extent possible, the reference conditions should be represented by actual
wetlands."
6 When choosing a reference standard, one needs to choose whether such a standard represents the Minimally
Disturbed Condition (MDC) or Least Disturbed Condition (LDC), or a combination of the two, based on best
attainable condition (BAG). Muggins and Dzialowski (2005) note that MDC and LDC set the high and low end of
what could be considered reference standard condition. They go on to say that "these two definitions can be used
to help define the Best Achievable Conditions (BAC's), which are conditions that are equivalent to LDC's where the
best possible management practices are in use. The MDC's and LDC's set the upper and lower limits of the BAC's.
Using the population distribution of measures of biological condition associated with a reference population might
provide some insights regarding the potential relationship between the MDC and LDC for a particular region."
23
-------
Reference gradients serve several purposes. First, they are the source of information on what
constitutes a characteristic and sustainable level of integrity across the suite of ecological attributes
selected for a type (i.e., by visiting ecosystems that range from excellent to degraded in integrity we can
document the characteristics of each level of condition). Second, reference gradients establish the
range and variability of conditions exhibited by assessment metrics and they provide the data necessary
for calibrating assessment variables and models (i.e., they help guide the interpretation of wetland
status and trends assessments). Finally, they provide a concrete physical representation of wetland
ecosystems that can be observed and re-measured as needed (Smith et al. 1995, Klimas et al. 2006).
INDICATOR RATINGS
With a well-chosen set of metrics selected to track changes in the major and key ecological attributes,
based on the natural range of variability, with thresholds established for each metrics, we can structure
the rating system to guide our interpretation of ecological integrity, using simple scorecard grades:
Excellent: The indicator lies well within its range of natural variability.
Good: The indicator lies within but is near to its range of natural variability.
Fair The indicator lies outside its range of natural variability, but not outside its threshold of
ecological collapse.
Poor: The indicator lies near to well outside its threshold of ecological collapse.
Indicators with higher levels of integrity would generally be rated "A", "B", or "C" (from "excellent to at
least "fair" integrity), and those with substantial degradation are rated "D" ("poor" integrity) (see Table
1).
INDICATORS AT MULTIPLE SCALES (LEVEL 1 TO LEVEL 3)
Overview of the 3 levels
The selection of metrics to assess ecological integrity can be executed at three levels of intensity
depending on the purpose and design of the data collection effort (Brooks et al. 2004, Tiner 2004, US
EPA 2006). This "3-level approach" to assessments, summarized in Table A3, allows the flexibility to
develop data for many sites that cannot readily be visited or intensively studied, permits more
widespread assessment, while still allowing for detailed monitoring data at selected sites.
24
-------
TABLE A3. Summary of 3-level approach to conducting ecological integrity assessments (adapted from
Brooks et al. 2004, USEPA 2006).
Level 1 - Remote Assessment
Level 2 - Rapid Assessment
Level 3 - Intensive Assessment
General description:
Imagery based assessment of
landscapes
General description:
Rapid site integrity assessment
General description:
Detailed site integrity assessment
Evaluates: Integrity of both on and off-
site conditions around individual
sites/occurrences using
indicators within occurrences that are
visible with remote sensing data, and
Indicators in the surrounding landscape /
watershed
Evaluates: Integrity of individual
areas/occurrences using relatively simple
field indicators
• Very rapid assessment (narrative)
• Rapid assessment (standard metrics)
• Hybrid assessments (rapid + vegetation
Plot)
Evaluates: Integrity of individual
areas/occurrences using relatively
detailed quantitative field indicators.
Choice of metrics for detailed
assessment may differ from that for
monitoring.
Based on:
• GIS and remote sensing data
• Layers typically include:
• Land cover, land use, other ecological
types
• Stressor metrics (e.g., roads, land use)
Based on:
• On-site condition metrics (e.g.,
vegetation, hydrology, soils,)
• Stressor metrics (e.g., ditching, road
crossings, and pollutant inputs)
Based on:
• On-site condition metrics (e.g.,
vegetation, hydrology, soils)
• Indicators that have been calibrated to
measure responses of the ecological
system to disturbances (e.g., indices of
biotic or ecological integrity)
Potential uses:
• Identifies priority sites
• Identifies status and trends of acreages
across the landscape
• Identifies condition of ecological types
across the landscape
• Informs targeted restoration and
monitoring
Potential uses:
• Relatively inexpensive field
observations across many sites
• Informs monitoring for implementation
of restoration, mitigation or
management projects
• Landscape / watershed planning
• General conservation and management
planning
Potential uses:
• Detailed field observations, with
repeatable measurements, and
statistical sampling design
• Identifies status and trends of specific
occurrences or indicators
• Informs monitoring for restoration,
mitigation, and management projects
Level 1 Remote Assessments rely almost entirely on Geographic Information Systems (GIS) and remote
sensing data to obtain information about landscape condition and stressors in and around an
occurrence. They can also help assess the distribution and abundance of ecological types in the
landscape or watershed. Level 2 Rapid Assessments use relatively simple field metrics for collecting
data on specific occurrences, and will often require considerable professional judgment. Our approach
emphasizes a condition-based rapid assessment, supplemented by information on stressors that may be
affecting condition. Level 3 Intensive Assessments require more rigorous, field-based methods that
25
-------
provide higher-resolution information on the wetland occurring within an assessment area, often
employing quantitative plot-based assessment procedures coupled with a sampling design. Calculations
of calibrated indices, such as a Vegetation Index of Biological Integrity (VIBI) may also be used. This 3-
level approach to assessments, summarized in Table 3, allows the flexibility for developing data on many
occurrences that cannot readily be visited or intensively studied as well as those for which detailed
information is desirable. When coupled with standardized procedures for defining occurrences across
the landscape (NatureServe generic EO specs), it encourages a widespread application of ecological
integrity assessments (assigning EO ranks) based on a reasonable and cost-effective approach for the
programmatic or project needs.
The 3-level approach is intended to provide increasing accuracy of ecological integrity assessment,
recognizing that not all conservation and management decisions need equal levels of accuracy. At the
same time, the 3-level approach allows users to choose their assessment based in part on the level of
classification (and thereby the specificity of the conceptual model).
To ensure that the 3-level approach is consistent in how ecological integrity is assessed among levels, a
standard framework or conceptual model, such as the EIA Model introduced above, should be used for
choosing metrics. Using this model, a similar set of metrics would be chosen across the 3 levels,
organized by the standard set of ecological attributes and factors, such as landscape context, size, and
condition.
Calibrating the 3-Level Approach
Ideally, information at the three levels of assessment provides relatively consistent information about
ecological integrity, with improved interpretations as the level of intensity goes up. To achieve this, the
various levels need to be calibrated against each other. For example, sites where a Level 2 or Level 3 IEI
or VIBI has been determined can be used to calibrate the Level 1 remote-sensing based index of integrity
(Mack 2006, Mita et al. 2007, Fennessy et al. 2007).
LEVEL 1 ASSESSMENT (REMOTE-SENSING METRICS)
Overview
Level 1 Assessments are based primarily on metrics derived from remote sensing imagery. The goal is to
develop metrics that assess the landscape context and the on-site conditions of an ecosystem. Satellite
imagery and aerial photos are the most common sources of information for these assessments.
Typically it is the stressors to the ecological integrity of ecosystems that are most observable with these
sources of information, so condition is evaluated through the lens of stressors. Level 1 assessments are
widely used as part of regional assessments because of their ability to characterize large landscape
areas.
There are growing sets of information on various kinds of stressors that impact ecosystems. Danz et al.
(2007) noted that "Integrated, quantitative expressions of anthropogenic stress over large geographic
regions can be valuable tools in environmental research and management." When they take the form
of a map, or spatial model, these tools initially characterize ecological conditions on the ground; from
highly disturbed to apparently unaltered conditions. They can be particularly helpful for screening
candidate reference sites; i.e., a set of sites where anthropogenic stressors range from low to high.
Ecological condition of reference sites are further characterized to determine how ecological attributes
are responding to apparent stressors. This knowledge may then apply in other similar sites.
26
-------
Anthropogenic stressors come in many forms, from regional patterns of acid deposition or climate-
induced ecosystem change, to local-scale patterns in agricultural drainage ditches and tiles, point-source
pollution, land-conversion, and transportation corridors, among others. To be effective, a landscape
condition model needs to incorporate multiple stressors, their varying individual intensities, the
combined and cumulative effect of those stressors, and if possible, some measure of distance away from
each stressor where negative effects remain likely. Since our knowledge of natural ecosystems is varied
and often limited, a primary challenge is to identify those stressors that likely have the most degrading
effects on ecosystems or species of interest. A second challenge is to acquire mapped information that
realistically portrays those stressors. In addition, there are tradeoffs in costs, complexity, the often
varying spatial resolutions in available maps, and the variable ways stressors operate across diverse land
and waterscapes. Typically, expert knowledge forms the basis of stressor selection, and relative
weighting. Once models are developed, they may be calibrated with field measurements. Developing
empirical relationships between stress variables and ecological response variables is a key to providing
insights into how human activities impact ecological condition (Danz et al. 2007).
Two related approaches may be taken to developing Level 1 metrics. First, emphasis can be placed on a
comprehensive evaluation of the entire landscape, based on mapped information of stressors. The
method compiles and integrates multiple layers of information into an overall synthetic index of
landscape and site stressors (as e.g. described by Danz et al. 2007, also Mack 2006) into an overall. The
overall index can also be decomposed into individual stressors or sets of stressors, to determine which
may be most important. This is a "stressor-based approach," and it assesses ecological integrity of
occurrences at specific sites somewhat indirectly.
Second, emphasis can be placed on a method where sites are evaluated using remotes sensing metrics
that estimate ecological integrity more directly. The overall index brings together a series of metrics
from site, buffer and surrounding landscape that characterize ecological integrity.
For example, NatureServe has developed a Landscape Condition Model (LCM, Comer and Hak 2009),
similar to the Landscape Development Index used by Mack (2006). It is a regional GIS model of
landscape condition, originally established as a 30m grid of unique values. The algorithm integrates
various land use GIS layers (roads, land cover, water diversions, groundwater wells, dams, mines, etc.)
(Table 1). These layers are the basis for various stressor-based metrics. The metrics are weighted
according to their perceived impact on ecological integrity, into a distance-based, decay function to
determine what effect these stressors have on landscape integrity. The result is that each grid-cell (30
m) is assigned a "score". The product is a watershed map depicting areas according to their potential
"integrity." The index is segmented into three or four rank classes, from Excellent (minimally disturbed)
(A) to Poor (degraded) (D), in accordance with Table A4.
27
-------
TABLE A4. Example of Level 1 metrics for assessing ecological integrity of
an ecosystem, based primarily on "stressor metrics."
Transportation 1
Secondary and connecting roads
Theme
Primary Highways with limited access
Primary Highways without limited access
Local, neighborhood and connecting roads
Urban and Industrial Development
High Density Developed
Medium Density Development
Low Density Development
Managed & Modified Land Cover
Cultivated Agriculture
Pasture & Hay
Managed Tree Plantations
Introduced Upland Herbaceous
Introduced Wetland Vegetation
Introduced Tree & Shrub
Recently Logged
Native Vegetation with Introduced Species
Ruderal Forest & Upland
1 A common source of data for these stressors in the U.S. are the ESRI®
Data & Maps: StreetMap™ Series issue: 2006 United States, 1:100,000
and National Land Cover Data/ LANDFIRE Existing Vegetation. 2001-
2003 United States. 30m pixel/ 1:100,000
NatureServe has also developed a number of Level 1 EIA methods, which uses some of the same
information available in the Landscape Condition Model, but selecting measures that are more
directly relevant as indicators of integrity, such as buffer extent or connectivity. One version,
shown in Table A5, uses fairly simple measures from remote sensing imagery to develop relatively
simple level 1 metrics (what can be called "tier 1" metrics). For example, connectivity is measured
simply as the proportion of natural land cover in the landscape area around an occurrence. This
version of Level 1 EIA is general and may be applicable to all terrestrial ecosystems.
28
-------
TABLE A5. Example of a general multi-metric approach
fora Level 1 Ecological Integrity Assessment.
RANK FACTOR
Metrics
Submetrics
Weight
LANDSCAPE CONTEXT
Connectivity
0.5
Connectivity: % Natural Land Cover in 100 ha area
Connectivity: % Natural Land Cover in 1000 ha area
Surrounding Land Use Index
0.5
Surrounding Land Use: Score for 100 ha area
Surrounding Land Use: Score for 1000 ha area
Buffer Index
Percent Buffer
Average Buffer Width
SIZE
Size
CONDITION
On-Site Land Use Index
More sophisticated Level 1 metrics can be developed, where the imagery is interpreted in greater
detail (we can refer to these as tier 2 metrics) (Table A6). For example, connectivity can be
assessed using "Circuitscape" which uses circuit theory to predict connectivity in heterogeneous
landscapes for individual movement, gene flow, and conservation planning (see
http://www.circuitscape.org/Circuitscape/Welcome.html). Landscapes are represented as
conductive surfaces, with low resistances assigned to habitats that are most permeable to
ecological processes such as species movement or best promote gene flow, and high resistances
assigned to poor dispersal habitat or to movement barriers (McRae et al. 2008). Similar other
metrics, such as Fire Regime, are estimated from information on age class distributions modeled on
the landscape. Because these models draw on more detailed information, the metrics can be
tailored to more specific sets of ecosystems. For example, the fire regime class metric may have
different ratings for systems with broadly different fire regimes. (See full details in Appendix 1).
29
-------
TABLE A6. Example of a specific multi-metric approach for a Level 1 Ecological Integrity Assessment
for Great Basin Dryland Ecosystems. A full example is provided in Appendix A.5.1.
Rank Factor
Key Ecological Attribute and Specific Indicator
LANDSCAPE CONTEXT
Landscape Connectivity:-CircuitScape Index
Landscape Condition: Landscape Condition Model Index
SIZE
Change in Extent: Relative Area Lost to Land Conversion
CONDITION
Fire Regime: Departure from expected distribution of age classes using SCLASS
Native Species Composition: Invasive Annual Cover
The Role of Level 1 Assessments for Field-based Surveys
The Level 1 integrity ranks are often used as a means of prioritizing sites for field visits, where Level
2 or Level 3 assessments will be completed (e.g., see Fennessy et al. 2007), and ranks based on
those assessments would supersede these ranks. Thus level lassessments can be informative
about the overall range in conditions across a population of wetlands in a landscape or region.
They can serve as a helpful screening method for identifying the most likely conditions on the
ground.
Level 1 ratings can also be used as predictors of Level 2 or 3 ratings at individual sites. Tests
completed to date, however, show that Level 1 methods do not accurately predict individual site
ratings, particularly on-site conditions (Mack 2006, Fennessy et al. 2007). Our recent tests for
wetland site in Michigan and Indiana bear this out (Fig. A4). However, the models are more
successful in predicting overall IEI scores, because landscape context and size, as well as on-site
condition are relevant to an IEI. It may also be possible to re-calibrate the metrics used for Level 1
assessments based on these Level 2 scores.
30
-------
FIGURE A4. Examples of a correlation between two kinds of Level 1 EIA (remote sensing) models
and Level 2 (field-based) assessments of On-site Condition (with scale of A (5.0) to D (D= 1.25).
Condition ratings integrate field-based vegetation, soils, and hydrology metrics. The Landscape
Condition Model (LCM) uses stressor-based metrics and the Level 1 EIA Method has a combination
of stressor and integrity based metrics.
FIGURE A4a) LCM rating within the combined area of wetland and core landscape (1 km buffer
around wetland). Kruskal-Wallis F = 32.6, p < 0.001. VL= L< M< H =VH. (from Faber-Langendoen et
al. 2011).
Very High High
Medium Low Very Low
Landscape Condition Model Rating
FIGURE A4b) Level 1 EIA Landscape Context ratings, based on three landscape and buffer metrics.
Kruskal-Wallis F = 52.0, p < 0.001. A>B» C»D (Faber-Langendoen et al. 2011)
I .,
B <•>
§
O
C D
Landscape Context Rating
31
-------
LEVEL 2 ASSESSMENT (RAPID FIELD-BASED METRICS)
Level 2 Ecological Integrity Assessment metrics
The intent of ecological-integrity-based rapid assessment methods (RAMs) is to evaluate the
complex ecological condition of a selected ecosystem using a specific set of observable field
indicators, and to express the relative integrity of a particular occurrence in a manner that informs
decision-making, whether for restoration, mitigation, conservation planning, or other ecosystem
management goals (Stein et al. 2009). These Level 2 assessments are structured tools combining
scientific understanding of ecosystem structure, composition, and processes with best professional
judgment in a consistent, systematic, and repeatable manner (Sutula et al. 2006).
Metrics that are chosen should informative about integrity or sustainability of major or key
ecological attributes and to associated stressors (this is sometimes described as the metrics
showing a "stressor-dose response" to changes in stressor levels). Stressor tests can be conducted
to ensure that metrics are informative, by assessing how metrics respond to a gradient of stressors
levels (Rocchio 2007, Jacobs et al. 2010, Faber-Langendoen et al. 2011).
Level 2 assessments rely primarily on relatively rapid (~2- 4 hour) field-based site visits, but this
may vary, depending on the purposes of the assessment. They provide the opportunity to do
direct, ground based surveys of ecosystem occurrences. RAMs have particularly widely available
for wetlands because of the need for mitigation and restoration tools, and they are in use by many
state wetland programs (Fennessy et al. 2007). Typically three to five metrics are identified for
each of the ecological factors, each metric relevant to a key ecological attribute.
Examples of the range of metrics that can be completed for rapid assessments for wetlands are
provided in Table A7 and A8. NatureServe has developed a Level 2 Ecological Integrity Assessment
method for all wetlands in the U.S., with some metrics having variants for certain ecosystem types
(using formations and macrogroups) or hydrogeomorphic types (using HGM classes) (see Table A7)
(Faber-Langendoen et al. 2008, Faber-Langendoen 2009). EPA has also developed a rapid
assessment tool as part of an ambient monitoring program, called USA RAM (Collins and Fennessy
2010 draft) (Table A8). The protocol is similar to that of the NatureServe EIA method, but metrics
more often measure degree of complexity or diversity of ecosystems, which may or may not be
reflective of ecological integrity.
32
-------
TABLE A7. Example of a standard set of indicators based on the conceptual model of Ecological
Integrity, for wetlands. Indicators occur at different levels of precision (rank factor, major
ecological factor, metric). Not shown are how some of the metrics have variants based on
wetland type (bog & fen, marsh, floodplain & swamp forests, mangrove, etc.). See Appendix 2
for details on each metric.
RANK FACTOR
LANDSCAPE
CONTEXT
SIZE
CONDITION
MAJOR
ECOLOGICAL
FACTOR
LANDSCAPE
CONTEXT
BUFFER
SIZE
VEGETATION
HYDROLOGY
SOIL
METRIC NAME
Connectivity (Core, Supporting)
Land Use Index (Core, Supporting)
Barriers to Landward Migration (% of Perimeter
Obstructed) (tidal)
Buffer Index
• Average Buffer Width
• Percent Buffer
• Buffer Condition
Relative Patch Size (ha)
Absolute Patch Size (ha)
Vegetation Structure
Regeneration (woody)
Native Plant Species- Cover
Invasive Exotic Plant Species - Cover
Vegetation Composition
Water Source
Hydroperiod
Hydrologic Connectivity
Physical Patch Types
Soil Disturbance (Soil Surface Condition)
The checklists provide additional field information on stressors to the wetland site or occurrence.
Details are provided in Faber-Langendoen et al. (2008).
33
-------
TABLE A8. Draft Scoring Criteria for USA-RAM vlO.
ECOLOGICAL FACTOR
LANDSCAPE
BUFFER
HYDROLOGY
PHYSICAL STRUCTURE
VEGETATION
METRIC NAME
Wetland Abundance:
Landscape Connectivity
Percent of Assessment Area having Buffer
Buffer Width
Hydroperiod
Hydrologic Connectivity
Topographic Complexity
Patch Type Diversity
Vertical Complexity
Plant Community Complexity
Level 2 Stressors Checklists
Stressor checklists can be useful as additional information when evaluating the ecological integrity
of an occurrence (Kapos et al. 2002). Typically, they are an aid to further understanding the overall
condition of the wetland. In some cases, where stressors appear to be having a negative impact on
the site, but the condition metrics do not reflect these impacts, it may lead to changes in the
overall index of ecological integrity of a wetland. This should only be done in exceptional
circumstances. The need for manual over-rides may suggest that the current condition metrics may
be insensitive to degradation due to certain stressors, and future adjustments to the metrics used
may be needed.
Stressors are listed if they are observed or inferred to occur, but are not included if they are
projected to occur in the near term, but do not yet occur. Stressors may be characterized in terms
of scope and severity. Scope is defined as the proportion of the occurrence of an ecosystem that
can reasonably be expected to be affected (that is, subject to one or more stresses) by the threat
with continuation of current circumstances and trends. Within the scope (as defined spatially and
temporally in assessing the scope of the threat), severity is the level of damage to the ecosystem
from the threat that can reasonably be expected with continuation of current circumstances and
trends by excluding potential new threats). For ecosystems, severity is typically assessed by known
or inferred degree of degradation or decline in integrity to one or more key ecological attributes.
Standardized checklists of stressors have been developed for a variety of rapid assessment
methods (Collins et al. 2006, Faber-Langendoen et al. 2008, Collins and Fennessy 2010). They can
be used to create field-based versions of stressor indices, e.g., the Human Stressor Index of Rocchio
(2007) integrates stressor scores for hydrology, soils, and buffer.
Variation on the Level 2 Assessment
It is worth noting several variants of the Level 2 EIA assessment methods that may appeal to
different needs. First, there is the "very rapid assessment method," in which, the attributes
themselves serve as the general indicators, and field crews complete a structured narrative
evaluation of those attributes. For example, field crews may record observations on the buffer,
vegetation, soils, and hydrology, and then rate the attribute directly. While not preferred as a
34
-------
general, standard method, it remains a valuable approach for professional ecologists, well-
experienced in the range of variation in wetland conditions and degradation, provided the reasons
being the ratings are documented. It can also be a form of accuracy assessment for Level 1
assessments. This approach has been widely used by the Natural Heritage Network.
A second variant is to complete a standard level 2 method, but add a few select level 3 indicators
because it is important for the goals of the project to better understand some key attributes. It
may also be desirable to continue collecting detailed information on certain attributes to validate
the level 2 assessment. We may refer to these as the "slower rapid assessment method." A
common addition is that of a vegetation plot, or some type of standardized plant species list for an
occurrence. These data can provide sufficient composition information for VIBIs or FQIs, or
structural characteristics (e.g., old growth or coarse woody debris ratings in forests). As long as the
modifications are structured within the overall conceptual model, there should be little difficulty in
producing comparable results to other RAMs. This approach has also been widely used by the
Natural Heritage Network.
Validation of Level 2 Methods
Because RAMs are rapid-based metrics and rely in part on best professional judgment, it is
important that they be calibrated and validated against independent measures of wetland
condition in order to establish their scientific defensibility (Sutula et al. 2006, Fennessy et al. 2007).
But RAMS often assess a wider range of ecological attributes of wetland integrity (from buffer to
hydrol, vegetation and soils, as the EIA conceptual model calls for, so the challenge is to identify a
comprehensive set of intensive metrics that span the same range of attributes. Otherwise,
calibrations run the risk of optimizing the RAM for only one or several attributes. For example,
many level 3 assessments focus on vegetation or biota (e.g., VIBIs), expecting in part that they
integrate the response expected from the rest of the biophysical environment. Although this may
be true, it too comes with assumptions. Thus, a truly comparable set of level 3 assessments are
needed to calibrate level 2 assessments. That said, level 3 assessments of particular attributes can
help validate parts of a level 2 assessment. As Stein et al. (2009) point out, decisions regarding
modification of RAM components can be made based on a "weight-of-evidence" approach, that is,
to combine information from multiple lines of evidence to reach a conclusion. In fact, such an
approach is used to select indicators for RAMs in the first place.
LEVEL 3 ASSESSMENTS (INTENSIVE FIELD METRICS)
The intent of intensive methods for evaluating ecological-integrity is to develop data that are
rigorously collected, often with an explicit sampling design, to provide better opportunities to
assess trends in ecological integrity over time. The quantitative aspect of the indicators lends
themselves to more rigorous testing of the criteria for metric selection (see KEY ECOLOGICAL
ATTRIBUTES AND INDICATORS above). Because of their cost and complexity, level 3 methods are
often closely evaluated to ensure that they address key decision-making goals, whether for
restoration, mitigation, conservation planning, or other ecosystem management goals. They are
often highly structured methods, with detailed protocols that ensure a consistent, systematic, and
repeatable method (Sutula et al. 2006). The level of intensity required of level 3 methods typically
means that they are used in conjunction with level 1 and 2 methods to increase spatial
representation and maintain affordability.
As with other levels, metrics that are chosen should be informative about integrity or sustainability
of major or key ecological attributes and to associated stressors. Stressor tests can be conducted
35
-------
by assessing how metrics respond to a gradient of stressors levels (Rocchio 2007, Jacobs et al.
2010, Faber-Langendoen et al. 2011).
Level 3 metrics, more so than level 1 and 2, allow for greater specification by ecosystem type. The
detailed measures may allow for greater sensitivities in differences among ecosystems in terms of
ecological processes, structure or composition. Examples of the range of metrics that can be
completed for intensive assessments of forest ecosystems is provided in Appendix 3.
Some intensive assessments have focused on one major attribute, that of vegetation. As with
aquatic IBI methods (Karr and others), the approach has been to develop a Vegetation Index of
Biotic Integrity (VIBI). Use of plants for a terrestrial (dryland, wetland) IBI makes sense (Mack
2007): plants, and especially vascular plants are large, obvious, important components of terrestrial
ecosystems; their taxonomy is relatively well understood and regional and state-specific taxonomic
treatments are available; the flora is large and offers numerous potential attributes for the
development of a plant IBI; quantitative vegetation sampling methods are well developed and
relatively easy to implement in the field, sampling is cost-effective and the data sets acquired from
such sampling have multiple uses including IBI development, setting mitigation wetland
performance standards and supporting wetland permit program decision-making (Fennessy et al.
2002). An example of the individual metrics and descriptions that we will use to develop the VIBI
are shown in Table A9.
TABLE A9. Example of a VIBI for Freshwater Wet Meadows and Marshes (Mack 2007,Appendix A).
Description of metrics used VIBI-EMERGENT
Metric
Type
Description
_
Carex
Dicot
Shrub
Hydrophyte
A/P ratio
FQAI
% Sensitive
% Tolerant
% Invasive
graminoids
Biomass
Richness
Richness
Richness
Richness
Richness ratio
Weighted richness
index
Dominance ratio
Dominance ratio
Dominance ratio
Primary production
Number of species in the genus Carex. Note number Cyperaceae
species used as a substitute metric for Lake Erie Coastal Marshes
Number of native dicot (dicotyledon) species
Number of shrub species that are native and wetland (FACW,
OBL) species
Number of vascular plant species with Facultative Wet (FACW) or
Obligate (OBL) wetland indicator status (Andreas et al., 2004)
Ratio of number species with annual life cycles to number of
species with perennial life cycles. Biennial and woody species
excluded from calculation
The Floristic Quality Assessment Index score calculated using Eq.
(7) and the coefficients of Andreas et al. (2004)
Sum of relative cover of plants in herb and shrub stratums with a
coefficient of conservatism (C of C) of 6, 7, 8, 9 and 10 (Andreas et
al., 2004)
Sum of relative cover of plants in herb and shrub stratums with a
C of C of 0,1 and 2 (Andreas et al., 2004)
Sum of relative cover of Typha spp., Phalaris arundinacea, and
Phragmites australis
The average grams per square meter of clip plot samples
collected at each emergent wetland
36
-------
FIELD METHODS AND PROTOCOLS
Any discussion of metrics would be incomplete without at least a note that documentation of the
rationale and protocols for metrics is vital for their consistent application. The protocols guide the
field methods and data collection. They ensure that users understand how the assessment was
done, and in the case of monitoring programs, are able to repeat the measurements. See Oakley
et al. (2003) for recommendations on content for metric protocols. Examples of metric protocols
are available in Faber-Langendoen et al. (2008), Tierney et al. (2010), and in Appendix 6. A data
dictionary of ecological integrity metrics is under development.
ECOLOGICAL INTEGRITY SCORE CARDS
When reporting on the results of assessments, a single index is often desired because it provides an
overall measure of the ecological integrity of the ecosystem. To be effective, it needs to be a)
grounded in the overall ecological integrity model so that the ecological information is clearly
summarized, b) be readily understood by managers and the public, and c) be helpful for decision-
making, such as whether wetlands are meeting water quality standards (Fennessy et al., 2007,
Jacobs et al. 2010). Having used multiple indicators to get a clear picture of the status of a key or
major attribute, it makes sense to use the weight of the evidence across indicators to determine
the status of the ecosystem.
Andreasen et al. (2001) outline six characteristics that a practical index of ecological integrity
should have:
• Multi-scaled
• Grounded in natural history
• Relevant and helpful (to the public and decision-makers, not just scientists)
• Flexible
• Measurable
• Comprehensive (for composition, structure and function).
We have previously outlined how our method addresses the six characteristics (Faber-Langendoen
et al. (2006). The construction of our ecological integrity assessments has been driven by the need
to represent major and key attributes of integrity, for which specific indicators or metrics are
identified. Thus one key feature of the scorecard is that the integrity of those attributes be
summarized through the information available from the metrics. Plus, all of the levels of
assessment that we describe are gathering data at the level of ecological factors, so scorecards will
also retain a common set of levels for reporting, a desirable feature of scorecards highlighted by
Harwell et al. (1999). This will also make it easy for users to find the specific indicator information
on which these summary scores are based.
The IEI reports on the condition of a wetland by estimating the degree to which individual sites
have departed from reference standard conditions. The specific indicators and the attributes are
scaled based on best understanding of the range of natural and other variability relevant to
ecological integrity, and with reference to sites where the highest scores reflect reference standard
conditions and the lowest scores representing highly disturbed sites (Stoddard et al. 2006). A
scorecard approach depends on a consistent scaling of the indicators or metrics, such that their
ratings are comparable with respect to levels of integrity. It is then reasonable to summarize the
37
-------
metric ratings and roll them into aggregate scores, including an overall Index of Ecological Integrity,
based on a weight of evidence approach (Linkov et al. 2009).
It may also be desirable to summarize the level of stressors present on the site, in order to indicate
what might be driving the current levels of integrity. In many cases, these data may only provide
correlating support for the level of integrity, and further studies, or supporting evidence from other
studies, will be needed to demonstrate causation.
The degree to which the scorecard might need to be customized based on several aspects of
ecosystems. First, the relative importance of attributes may differ among system types. For
example, hydrology typically plays a very strong role in wetlands, but a minor one in drylands;
within wetlands, buffer may play a larger role in depressional wetlands than in riverine systems
(Jacobs et al. 2010). Or size may be relatively unimportant to wetland types, such as seeps, that
only ever occur in small patches. Second, and related to the above, failure of certain attributes
may lead to the overall collapse of a system. Thus more weight can be given to poor ratings of
those key attributes. Here again, ecosystems classifications play a valuable role in providing
guidance on our understanding of the role of ecological attributes among systems, and can ensure
standardized evaluations whenever that ecosystem is encountered. Nonetheless, it may be
tempting to focus on the individuality of ecosystems at the expense of readily interpretable results;
customizing should be done only where strong and range-wide evidence that it improvise the
discriminating ability of the index, to ensure that scores for sites can be readily compared across
watershed, landscapes and regions.
There are a number of approaches to aggregating metrics, but the most common is the rather
simple non-interaction point-based approach, where each metric is scored and treated
independently. The point-based approach is consistent with that of many IBI scoring methods (e.g.
Karr and Chu 1999). The scorecard is structured using the conceptual model (Fig. A3 above). Each
metric within an ecological factor or attribute is assigned a weight, based on its perceived
importance. Ratings for each metric are presented, along with conversion to a point value for that
rating (e.g., A = 5 points, B = 4, C=3, D=2, E =1) (or A=4, B=3, C=2 and D=l). The points are
multiplied by the weight to get a score for the metric. The scores (weighted points) for all metrics
within a major attribute are summed and divided by the sum of the weights to get an attribute
score. Each major factor is also weighted (e.g., in wetlands, soils are often weighted less than either
hydrology or vegetation). The factor scores can be further aggregated by major rank factors
(landscape context, size, and condition). Finally their scores can be weighted and summed to get
an overall score, which is converted to an Index of Ecological Integrity. If desired, Vegetation, Soils
and Hydrology can be combined separately into a Condition score before producing an overall
index rating. An example of a summary scorecard is shown in Table A10. More detailed examples
are provided in Appendices 1, 4 and 5.
38
-------
TABLE A10. Example of an Ecological Integrity Scorecard.
RANK FACTOR
MAJOR ECOLOGICAL
FACTOR
/ Metric
ECOLOGICAL INTEGRITY
LANDSCAPE CONTEXT
LANDSCAPE
Connectivity
Land Use Index
BUFFER
Buffer Index
SIZE
SIZE
Relative Patch Size (ha)
Absolute Patch Size
CONDITION
VEGETATION
Vegetation Structure
Regeneration (woody)
Native Plants - Cover
Invasive Exotic Plants - Cover
Increasers - Cover
Vegetation Composition
HYDROLOGY
Water Source
Hydroperiod
Hydrologic Connectivity
SOIL
Physical Patch Types
Soil Disturbance
Rating
B
B
B
A
B
B
B
A
A
B
A
B
B
C
c
B
C
B
B
C
C
c
B
B
B
B
39
-------
A.7 DEFINITION OF ECOLOGICAL INTEGRITY RATING VALUES (A-D)
DEFINITIONS
The ecological integrity scorecard will bring together our understanding of current status of
ecological attributes, and include threshold values for both the best conceivable occurrences and
those having only fair viability or integrity (NatureServe 2002). To ensure that the final ratings of
ecological integrity have consistency wherever they are used, we provide a narrative summary of
the different levels of integrity. Thus, the integration of individual metrics into overall ratings
should help provide a perspective on ecological integrity consistent with these definitions (Table
All).
We offer these definitions partly to provide a global perspective on ecological integrity. This means
that the best occurrence in a particular jurisdiction or geographic area (e.g., ecoregion) may not be
highly ranked or even viable. Information about local prioritization of EOs can be recorded in
optional fields or existing comment fields.
The A through D rating presumes that a particular type is still recognizable at some level as "the
type/' despite varying levels of degradation. At some point, a degraded type will "cross the line"
(or be "transformed" in the words of SER 2004) into a separate, typically semi-natural or cultural
type. In some state-and-transition models these may be treated as shifts to an 'alternative state."
As a matter of practicality, the current system has been lost. This requires working with a set of
diagnostic classification criteria, based on composition, structure, and habitat.
RECAP OF NATURAL VARIABILITY AND ECOLOGICAL INTEGRITY
A few final observations are in order regarding the role of natural and historic variability in
informing these definitions of ecological integrity. These are provided to highlight both the
importance and limits of using the range of natural or historic variability.
An "A" rank need not be comparable to historical conditions. For example, bison in native Great
Plains prairies will not conceivably exist again in their historical condition with herds numbering in
the millions, but nevertheless a range of prairie occurrences with e.g., managed herds of differing
sizes and conditions, might still be reasonably achievable. In other words, it is still necessary to
conceive of a range of integrity, although the range is truncated when compared to EO rank
specifications that would have been written 150 years ago. (NatureServe 2002)
The "A" rank threshold should not be based solely on historical information because: a) historical
status often cannot be achievable; b) use of historical information could drastically truncate the
rank scale for current EOs; and c) historical information is often not known (NatureServe 2002).
None-the-less, for ecosystems, the A-ranked threshold should be based on a "minimally-disturbed"
reference state, whenever possible (Stoddard et al. 2006).
In order to set a threshold that is reasonably and conceivably achievable for 'TV-ranked
occurrences, it is necessary to consider restorability so that the threshold is not limited to EOs that
are extant (NatureServe 2002).
40
-------
TABLE A11. Definition of Index of Ecological Integrity values
(Faber-Langendoen et al. 2009c).
Rank Value
Description
Occurrence is believed to be, across the range of a type, among the highest quality examples
with respect to key ecological attributes functioning within the bounds of natural disturbance
regimes. Characteristics include: the landscape context contains natural habitats that are
essentially unfragmented (reflective of intact ecological processes) and with little to no stressors;
the size is very large or much larger than the minimum dynamic area ; vegetation structure and
composition, soil status, and hydrological function are well within natural ranges of variation,
exotics (non-natives) are essentially absent or have negligible negative impact; and, a
comprehensive set of key plant and animal indicators are present.
Occurrence is not among the highest quality examples, but nevertheless exhibits favorable
characteristics with respect to key ecological attributes functioning within the bounds of natural
disturbance regimes. Characteristics include: the landscape context contains largely natural
habitats that are minimally fragmented with few stressors; the size is large or above the
minimum dynamic area, the vegetation structure and composition, soils, and hydrology are
functioning within natural ranges of variation; invasives and exotics (non-natives) are present in
only minor amounts, or have or minor negative impact; and many key plant and animal
indicators are present.
Occurrence has a number of unfavorable characteristics with respect to the key ecological
attributes, natural disturbance regimes. Characteristics include: the landscape context contains
natural habitat that is moderately fragmented, with several stressors; the size is small or below,
but near the minimum dynamic area; the vegetation structure and composition, soils, and
hydrology are altered somewhat outside their natural range of variation; invasives and exotics
(non-natives) may be a sizeable minority of the species abundance, or have moderately negative
impacts; and many key plant and animal indicators are absent. Some management is needed to
maintain or restore7 these key ecological attributes.
Occurrence has severely altered characteristics (but still meets minimum criteria for the type),
with respect to the key ecological attributes. Characteristics include: the landscape context
contains little natural habitat and is very fragmented; size is very small or well below the
minimum dynamic area; the vegetation structure and composition, soils, and hydrology are
severely altered well beyond their natural range of variation; invasives or exotics (non-natives)
exert a strong negative impact, and most, if not all, key plant and animal indicators are absent.
There may be little long-term conservation value without restoration, and such restoration may
be difficult or uncertain.8
7 By ecological restoration, we mean "the process of assisting the recovery of an ecosystem that has been
degraded, damaged, or destroyed... Restoration attempts to return an ecosystem to its historic trajectory" (SER
2004). Restoration may be distinct from rehabilitation, reclamation, creation, mitigation, or ecological
engineering, unless they have as part of their goal a restoration as defined above (see SER 2004 for details).
8 D-ranked sites present challenges. For example, with respect to classification, a degraded type may bear little
resemblance to examples in better condition. Whether a degraded type has "crossed the line" ("transformed"
in the words of SER 2004) into a new semi-natural or cultural type is a matter of classification criteria. Here
we include D ranked examples as still identifiable to the type based on sufficient diagnostic criteria present.
41
-------
A.8 RETURNING TO THE WATERSHED OR LANDSCAPE SCALE
A clear and consistent scorecard at the site level that is readily repeatable for occurrences
everywhere leads us back to the role of ecological integrity assessment methods at larger spatial
scales. The IEI can be used to report on the status of ecosystems across watersheds or landscapes
(Fig. A5). Jacobs et al. (2010) note how individual indicators for wetlands can easily be relayed to
the public or environmental managers to communicate the status of certain kinds of wetland types
in the Nanticoke River watershed. Similarly, Faber-Langendoen et al. (2011) use a scorecard
approach to report on the overall integrity of individual wetlands across an ecoregions (Fig. A6).
This information can then be used to direct management efforts in the appropriate areas to
improve the condition of types in a watershed.
FIGURE A5. Ecological Integrity Scorecard Roll-up for Central Basin and Range ecoregions in the
southwest United States.
Pinyon-Juniper Composite
28-0.78
42
-------
FIGURE A6. Ecological Integrity Scorecard Roll-up for ecoregions of Omernik Ecoregions in northern
Indiana and southern Michigan.
"
I
^ •
.
ftM
MM
a*
-'
'~n»w»
- ' ^
Appalachian
Interior Plateau and Pialrle Fen
Rounded EO Rank
• A
B
C
• D
B = iiateiankarB.B?.orBC
C = 3to<« ran* of C. C">. Cf CD
D 9 $tM« i:ni fi( r> D?
A.9 ECOLOGICAL INTEGRITY, CONSERVATION STATUS AND ECOSYSTEM
SERVICES
Ecological integrity is only one aspect of interest for ecosystem assessments. Two common ones
include: 1) conservation (at-risk) status of ecosystem types and biodiversity value, which includes
aspects of wetland irreplaceability, 2) functional values and ecosystem services (Hruby 2001,
Fennessy et al. 2004). The first aspect, assessing the conservation status and irreplaceability value
of ecosystem types and occurrences, can be part of a risk assessment process, where more
irreplaceable systems are preferentially targeted for threat abatement or subject to greater degree
of protection, thereby avoiding losses that lead to challenging mitigation or restoration efforts.
This assessment can begin by assessing the relative conservation status (or risk of extirpation) of a
given type. For example, the Heinz Center (2002) uses the "At-risk wetland plant communities"
(based on NatureServe's conservation status assessment approach), as an indicator of overall
wetland or aquatic condition.
The 2nd aspect, that of wetland functional or ecosystems services value, has been widely developed
as part of the functional assessments completed by the hydrogeomorphic (HGM) approach.
Functional assessments and ElAs differ in important respects (Table A12), even as they share many
field data methods. Functional assessments categorize wetland types by creating a number of
broad wetland classes based on hydrogeomorphic characteristics, then allowing regional
applications to specify subclasses. We suggest that the indicators and ecological attributes
developed for ecological integrity assessments will often be useful for functional assessments,
43
-------
though additional data may need to be collected. For example, in a functional assessment, a series
of measures (e.g., litter + O-horizon thickness + coarse woody debris + snags) are combined with
flooding frequency to estimate the degree to which a wetland exports organic carbon, whereas in
an ecological integrity assessment, these measures would be combined into abiotic and biotic
metrics that assess departures from the ranges of natural variability and characteristics of
benchmark sites for these ecosystems.
TABLE A12. Comparison of Ecological Integrity (Condition) and Functional Wetland Assessments.
See Fennessy et al. (2007) for further comparisons between the two approaches.
Purpose
"Currency"
Approach
Method
Application
Ecological Integrity/ Condition
Assessment
Estimate current ecological integrity
Condition of major and key ecological
attributes
"Holistic" ecological integrity
Combines indicators into conceptual
model of ecological factors and key and
key ecological attributes
Mitigation, monitoring, state water
quality standards, and Heritage
Network.
Functional Assessment
-Estimate ecological functions (HGM)
Level of functions and ecological
services
"Compartmental;" each function
assessed individually.
Combines indicators into conceptual
model of ecological functions and
values
Mitigation and monitoring.
HGM and other ecosystem services methods may also differ from ecological integrity assessments
in that they evaluate the level or capacity of wetland functions, rather than the status of key
ecological attributes. They may be concerned with the level or capacity of each function regardless
of how or whether it relates to ecological integrity. Thus, a wetland with excellent integrity will
perform all of its functions at levels expected for its wetland class or type, whether or not these are
optimal levels with respect to desired functions or ecosystem services. Many HGM functional
assessments collect very similar data to an ecological integrity assessment; what differs is that the
functional assessments may take these data and develop logical operators to infer function.
To enhance the collection of both ecological integrity and ecosystem services, measures can be
added to the field or remote sensing methods. For example, National Wetland Inventory maps
include a wetland type classification (Cowardin et al. 1985), and more recently include an NWI+
application, which describes the hydrologic functions of the watershed and site scales (USFWS
2010).
44
-------
A.10 ADAPTING THE ASSESSMENT OVER TIME
We conclude by noting that our efforts to assess ecological integrity are only approximations of our
current understanding of the system. Ecosystems are far too complex to be fully represented by a
suite of metrics and attributes. Moreover, our metrics, indices and scorecards must be flexible
enough to allow change over time as our knowledge grows. What is important is that we present
as clearly as we can how we are conducting our assessments, so that we foster communication and
understanding among people with different backgrounds, goals, and points of view.
NatureServe will upgrade its databases to manage and store the ecological assessments, including
the component metrics, and will encourage new version of metrics to be developed and
substituted for old ones as they become available. Programs and partners will be encouraged to
test and refine these metrics, keeping in mind the overall definitions and purposes of ecological
integrity assessments.
45
-------
SECTION B: IDENTIFYING A REFERENCE GRADIENT
OF WETLAND TYPE AND CONDITION
B.I INTRODUCTION
In this section we develop a sampling design that will help us identify a candidate set of wetlands
that span the ecological gradients in northern Indiana and southern Michigan (all of Omernik level
3 ecoregions 55, 56, and 57, and the Indiana part of 54). We use a sampling design based on
previously classified sites, remote sensing metrics and on-site evaluations to predict a reference
gradient of conditions (from minimally disturbed to degraded) for each wetland type. This design
step is critical, because we needed a sampling design that allowed us to test the relative sensitivity
of our Ecological Integrity Assessment (EIA) method to changes in ecological integrity across the full
range of wetland types, conditions and stressors (as detailed in Section C).
A follow-up application presented itself through the development of the sampling design. That is, if
we successfully create a sampling design that we hypothesize will span a range of conditions, and
we independently verify those conditions, then we can use that design to predict the reference
gradient for future studies. Predicting a reference gradient will be of great value for monitoring
and assessment programs (Fig. Bl). This is because, in selecting and establishing metrics for
assessing condition or ecological integrity, an assumption is made that some type of reference
condition can be defined; that is, it is possible to describe a series of states of wetland integrity,
from minimally disturbed to degraded (Stoddard et al. 2006). But there are challenges to
implementing the reference condition approach. First, one needs some means of establishing the
gradient of reference sites. Second, one needs to be able to identify sites where the range of
reference conditions are found; this can be problematic in regions where there is a long and
extensive history of human uses that have altered the landscape. Third, one needs to be able to
sample these sites in a timely and cost-effective manner to gather the data on reference condition.
For that reason, a feasible approach is needed to establish reference conditions and identify
candidate reference sites (Faber-Langendoen et al. 2009b).
46
-------
FIGURE Bl. A reference network of wetland sites where ecological integrity has been verified (left
panel), which serves to guide a sampling design to assess wetland conditions in a watershed or
region (right panel).
REFERENCE WETLANDS AND SAMPLING DESIGN
FOR WETLAND CONDITION ASSESSMENT
REFERENCE NETWORK
SAMPLE DESIGN
Our purpose in this section is to summarize our proposed sampling design for establishing a
reference gradient. Our level of inference for this design and the subsequent testing of our
ecological integrity methods will be the range of conditions across a set of occurrences of all major
wetland types in southern Michigan and northern Indiana. However, if successful, we further
believe that the range of types and conditions will be applicable across the temperate regions of
the U.S. and elsewhere.
B.2 SAMPLING DESIGN METHODOLOGY
PROJECT AREA
Our sampling area in southern Michigan and northern Indiana is based on the Omernik ecoregions,
which are also being used in the EPA National Wetland Condition Assessment (Fig. B2). We
conducted our sampling within a relatively similar set of adjacent ecoregions, 55, 56, and 57, with a
few samples from ecoregions 54 in northwest Indiana.
47
-------
FIGURE B2. Ecoregional framework that guides the project study area within
EPA Region 5. Shown are the Omernik ecoregions (Omernik 1987, updated
by USEPA 2007) that were sampled within Lower Michigan and northern
Indiana.
EPA Region 5 Ecoregions
EPA Level III Ecoregions
Level III Name
Central Corn Belt Plains
Eastern Corn Belt Plains
Huron/Erie Lake Plains
Southern Michigan/Northern Indiana Drift Plains
This project was coordinated by NatureServe, with field crews comprised of Michigan and Indiana
Heritage Program staff. EPA's National Wetland Condition Assessment (NWCA) Team provided
guidance and support.
CLASSIFICATION
The success of developing and assessing wetland ecological integrity depends on understanding the
structure, composition, and processes that govern the wide variety of ecosystem types. Ecological
classifications can be helpful tools in categorizing this variety. They help ecologists to better cope
with natural variability within and among types so that differences between occurrences with good
integrity and poor integrity can be more clearly recognized. Classifications are also important in
establishing "ecological equivalency/' i.e., ensuring that degraded examples of a type are compared
with minimally disturbed examples of the same type. We use a variety of classifications in order to
effectively address biotic and abiotic aspects of wetlands, at different cales, but our primary focus
is on the USNVC macrogroup and the more finely scaled State Natural Community Type. We link
these types to NatureServe's Ecological Systems and to Hydrogeomorphic Types.
-------
USNVC Macrogroup
In the United States, the U.S. National Vegetation Classification (NVC) is supported by the Federal
Geographic Data Committee (FGDC), NatureServe, and the Ecological Society of America, with
other partners (FGDC 2008, Faber-Langendoen et al. 2009a, Jennings et al. 2009). The NVC was
developed to classify both wetlands and uplands and identify types based on vegetation
composition and structure and associated ecological factors. At the highest level of the
classification, Formation Class, there are 8 broad classes, each with 7 nested hierarchical levels,
which permit resolution of types from broad-scale formations to fine-scale associations (Table Bl).
We use the Macrogroup level for our assessments, of which there are seven in our region (Table
B2).
TABLE Bl. The following table illustrates the eight levels of the USNVC hierarchy for a
Midwest prairie fen. Also shown is an example of how NVC is a complementary
classification with Ecological Systems (mid-scale) and Natural Community (finer-scale)
types.
USNVC Hierarchy
Upper Levels
Formation Class
Formation Subclass
Formation
Mid-Levels
Division
Macrogroup
Group
Lower Levels
Alliance
Association
Pilot NVCTypes
Low Shrubland & Grassland
Temperate & Boreal Shrubland & Grassland
Temperate & Boreal Bog & Fen
North American Bog & Fen
Appalachian, Interior Plateau and Prairie Fen
North-Central Appalachian & Interior Seepage Fen
Alkaline Fen System 1
j
Shrubby-cinquefoil / Fine-leaved Sedges Fen Alliance
Prairie Fen (Michigan); Fen (Indiana) j
1 1
Shrubby-cinquefoil / Sterile Sedge - Big Bluestem - Indian-
plantain Fen Vegetation
49
-------
TABLE B2. List of formations and macrogroups found in the study area, and number of study
sites found for each macrogroup. Colloquial names for macrogroups are provided in square
brackets.
NVC FORMATION
Macrogroup
SWAMP & FLOODED FOREST
Northern & Central Floodplain Forest (M029) [Floodplain Forest]
Northern & Central Swamp Forest (M030) [Swamp Forest]
FRESHWATER SHRUBLAND, WET MEADOW & MARSH
Eastern North American Wet Shrub, Meadow & Marsh (M069) [Wet Shrub, Meadow & Marsh]
Great Plains Wet Meadow, Wet Prairie & Marsh (M071) [Wet Prairie]
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow (M067) [Coastal Plain Pondshore]
BOG & FEN
Appalachian, Interior Plateau & Prairie Fen (M061) [Rich Fen]
North American Boreal Bog & Fen (M062) [Bog & Poor Fen]
State Natural Community Type
Many states throughout the eastern United States have developed natural community
classifications with a focus on inventory and conservation applications, including Indiana (Jacquart
et al. 2002, see also www.in.gov/dnr/nature preserve/4743.htm) and Michigan (Kost et al. 2007).
They rely on a suite of state-level ecological characteristics, such as vegetation physiognomy,
species composition, soil moisture, substrate, soil reaction, or topographic position, to identify the
type. State natural community types are very comparable to alliances and associations in the
USNVC (Table Bl). We include the state classifications by macrogroup, because states report their
wetland information based on these types, and they help summarize the range of variation of
wetland types within macrogroups.
OTHER CLASSIFICATIONS
Ecological Systems
A second, related classification approach to that of the USNVC is the Ecological Systems
classification9 (Comer et al. 2003). It can be used in conjunction with the USNVC, roughly
corresponding to the "group" level, and below the macrogroup level (Table Bl, Table B3).
9 Ecological Systems in the U.S. are a component of the International Terrestrial Ecological Systems
Classification (Comer et al. 2003).
50
-------
Ecological Systems provide a spatial-ecologic perspective on the relation of associations and
alliances (fine-scale plant community types), integrating vegetation with natural dynamics, soils,
hydrology, landscape setting, and other ecological processes. The Ecological Systems classification
facilitates mapping at meso-scales (1:24,000 - 1:100,000). Comprehensive Ecological Systems
maps are available across the country (Comer and Schulz 2007, www.landscope.org). We use
Ecological Systems maps to help identify where various macrogroups may be found across the
landscape, and to characterize the landscape surrounding wetlands.
TABLE B3. Macrogroups and Ecological Systems in the study area.
Macrogroup
Northern & Central Floodplain Forest
Northern & Central Swamp Forest
Atlantic and Gulf Coastal Plain Pondshore and
Wet Meadow
Eastern North America Wet Shrub, Meadow &
Marsh
Appalachian, Interior Plateau & Prairie Fen
North American Boreal Bog & Fen
Ecological System
North-Central Interior Floodplain
(Laurentian-Acadian Floodplain Forest)
Laurentian-Acadian Alkaline Conifer-Hardwood Swamp
North-Central Interior and Appalachian Rich Swamp
North-Central Interior Wet Flatwoods
Northern Atlantic Coastal Plain Pond
Laurentian-Acadian Freshwater Marsh
Laurentian-Acadian Wet Meadow-Shrub Swamp
Great Lakes Wet-Mesic Lakeplain Prairie
Northern Great Lakes Interdunal Wetland
Great Lakes Freshwater Estuary and Delta
Northern Great Lakes Coastal Marsh
North-Central Interior Shrub-Graminoid Alkaline Fen
Boreal-Laurentian Bog
Laurentian-Acadian Alkaline Fen
Boreal-Laurentian-Acadian Acidic Basin Fen
Hydrogeomorphic Classification
The hydrogeomorphic (HGM) classification developed by Brinson (1993) was developed in order to
assist the U.S. Army Corp of Engineers with the evaluation of wetland impacts. HGM identifies
groups of wetlands that function similarly, based on three fundamental factors: geomorphic
setting, water source, and hydrodynamics (Smith et al. 1995). HGM classifications are widely used
by wetland scientists. For each wetland occurrence that we visited, we assigned the HGM class.
51
-------
WETLAND OCCURRENCE DATA - NATURAL HERITAGE AND OTHER DATASETS
We first needed to establish our overall population of wetland sites. We assessed the following
sources of information:
• State Heritage site information, in which the wetland type (determined using state
type, NatureServe ecological system type, and NVC macrogroup type), location
(typically a point and a polygon), and condition evaluation, are available.
• All Michigan Rapid Assessment Method (MiRAM) sites where sufficient information
was available to determine macrogroup, and approximate condition, and
functional rating.
• The Ecological Systems map for the region, showing all major wetland sites,
classified by the NatureServe Ecological Systems types.
Heritage Datasets
The Natural Heritage programs in Michigan and Indiana have for many years been identifying high
quality (minimally disturbed) occurrences for all wetland types in the state (using state natural
community classifications), and for rarer types, a fuller range of sites. No other comparable
datasets are available in these states; in addition, many other states have Heritage Programs with
comparable data. Heritage programs have typically evaluated the condition of occurrences using
best professional judgment, with a minimal amount of quantitative information, assigning each
occurrence a scorecard grade, or element occurrence rank (EORANK): A (Excellent), B (Good), C
(Fair), and D (Poor). Programs may differ in how these grades or ranks are assigned, but typically
the primary focus was the on-site condition of the wetland.
The Heritage program databases contain the following core information on each occurrence:
• Site Name
• State Natural Community / NVC Association name
• Element Occurrence (EO) ID (unique database code)
• EO Rank
• Geo-coordinates
• Directions to Site
• Polygon delimiting the extent of the wetland type occurrence
• Size of polygon
• Ownership
• Date of visit(s)
• Occurrence Comments (general description of occurrence, including dominant species).
An example of the spatial information and some of the core fields of information is shown in Figure
B3.
52
-------
FIGURE B3. Atypical example of the kind of tabular and geospatial information
available from Heritage Programs. The hatched blue area shows a polygon of a
Southern Hardwood Swamp (state type): Northern & Central Swamp Forest
(NVC macrogroup), covering 39 ac (16 ha). The Element Occurrence ID (EO ID) is
3473, and the project ID is 74. Assigned rank was "B." Circular areas are
landscape areas evaluated as part of the landscape context, including buffer
(blue line), core landscape (red line) and supporting landscape (yellow).
Site: GARY LAKE MEADOW
NVC MG: NORTHERN & CENTRAL SWAMP FOREST
Ml TYPE: Southern Hardwood Swamp
Size: 39 ac (16 ha
Comment: Transitional to forested seep, wet meadow
Legend
* SteCentroid
• V*1land EO
_j Buffer-200m
] Core Landscape Area (100ha)
Supporting Landscape Area (lOOOha)
To compile this data into our framework, we crosswalked each state type to the USNVC
macrogroup level (Table Bl). State types are relatively fine-grained relative to macrogroups, so
state types typically nest cleanly with the NVC macrogroup; thus our confidence in classification of
sites at the macrogroup level is very high. We could then bring in the Heritage EORANK data, and
compile a list of wetland occurrences by macrogroup and Heritage EORANK across the several EPA
ecoregions of interest in southern Michigan and northern Indiana. More than 500 wetland
occurrences were found across the region.
MiRAM Sites
MiRAM assesses overall wetland function and condition at a wetland site, across multiple wetland
types (MDEQ 2008). However, the MiRAM score for the entire wetland is typically a good reflection
of the condition of an individual type within the bigger wetland (T. Losee pers. comm. 2009). Data
on the location of MiRAM sites was available as point coordinates (hardcopy maps showing
wetland polygons were also available).
MiRAM sites were assigned to a macrogroup based on descriptive text recorded for each site; in
our initial data analysis, we were unable to assign a macrogroup for 48 of 68 MiRAM sites due to
53
-------
insufficient descriptive information. Only those MiRAM sites with assigned macrogroups were
considered for the site selection.
To determine how best to use the MiRAM information on wetland condition, we compared
Michigan EO rankings to the MiRAM score for 18 MiRAM sites that are co-located with EOs.
MiRAM scores include a Floristic Quality Index (FOJ). The FOJ proved to be the best match to
Heritage EO rank. We found that the range of FQI values provided in Table B4 were the best means
to split those values into condition classes comparable to the Heritage EORANK; the match to the
EO rank was generally close, providing adequate means to assign MiRAM sites to both macrogroup
and condition ranks.
TABLE B4: Condition Ranges Used to Stratify Project Data.
Original Condition Ranking
Michigan EO Rank
A, AB, A?
B, BC, B?
C,C?
CD, D
X H ?
/\, i i, .
Indiana EO Rank
A, AB
B, BC, AC
C, BD
CD, D
X H ?
/\, i i, .
MiRAM FQI (all) Score
FQI > 45
35 < FQI < 45
25 < FQI < 35
FQI < 25
NA
Project Condition
Ranking
A
B
C
D
Not Used
Ecological Systems Maps
NatureServe has comprehensive Ecological Systems maps across the project area (and country)
(Comer and Schulz 2007, www.landscope.org). However, wetland types were typically mapped
very generally, above the macrogroup level, so its accuracy is not strong enough at fine-grained
scales for our purposes. These maps are better used at local catchment to watershed scales to
identify wetlands and types. Given this, we did not rely on the Systems maps for this project.
SAMPLING DESIGN FOR REFERENCE GRADIENT
Classification and Condition Strata
We used classification stratum and condition stratum for site selection and analysis. Our
classification stratum is the NVC wetland types at the macrogroup level. Our crosswalk from state
natural community types to the macrogroup level was very clean, so for most sites we have a high
confidence in the classification stratum.
Our second stratum is the condition stratum, which predicts the potential integrity or condition of
an occurrence. Here our confidence is not as strong, as Heritage EORANKS are sometimes
outdated, and are based on the best professional judgment of field ecologists who visited the site,
which may vary among ecologists and states. Nonetheless having pre-existing field observations is
very valuable. To increase the standardization of the grade, we combined the on-site EORANK with
54
-------
a remote-sensing based landscape-context evaluation. Three primary metrics were used:
naturalness of surrounding landscape, land uses within the landscape, and the extent and condition
of the buffer immediately surrounding the wetland (Table B5). Together they contribute to an
overall Landscape Context Rating (LC) rating. Details of the metrics are provided in Appendix 3.
TABLE B5. Landscape Context metrics used to develop a rating around each
Heritage occurrence. See also Figure B3.
Metric
Submetric
Weight
LANDSCAPE CONTEXT
Connectivity
0.5
Connectivity: % Natural Land Cover in core 100 ha area
Connectivity: % Natural Land Cover in supporting 1000 ha area
Surrounding Land Use Index
0.5
Surrounding Land Use: Score for core 100 ha area
Surrounding Land Use: Score for supporting 1000 ha area
Buffer Index
Percent Assessment Area with Buffer
Average Buffer Width
The overall score of these landscape metrics were then combined with the on-site condition
evaluations to assign each wetland occurrence to a condition stratum i.e., landscape context rating
+ on-site EORANK = condition stratum rating). We first combined the ratings for landscape context
and Heritage EORANK as follows:
Condition Stratum Design 1
A = both EORANK and LC = A.
B = EORANK and LC or LC and EORANK = A andr B, A and C, or B or B.
C = A and D, B and C, B and D, or C and C
D = Cand D, D and D.
We also later tested a less stringent version of this design (Condition Stratum 2) to see if future
applications might benefit from this version:
55
-------
Condition Stratum Design 2
A = both EORANK and LCI = A, or A or B.
B = EORANK and LC or LC and EORANK = A and C, B and B, or B and C.
C = A and D, B and D, or C and C, or C and D.
D = Dand D.
Site Selection and Sample Size
As described above, reference sites were chosen primarily from State Heritage element occurrence
records for wetland communities, supplemented as necessary with data from MiRAM. We
assigned a macrogroup and a condition rating based on the methods described for those two strata
above. Table B6 summarizes our achieved number of sampling sites by classification and condition
strata, which varies considerably from our initial target of 10 sites per cell. Although our goal was
to achieve a balanced a design to ensure that the testing of the EIA method spanned the full range
of wetland types and condition, neither the statistical analyses nor our overall interpretation
depend on having exactly 10 replicates. Thus, rather than engage in an effort to balance the cells,
we emphasized attaining 5 or more sites per cell. In addition, some types (e.g., Bog & Poor Fen) are
relatively rare and in difficult to access locations, and few degraded examples are available.
TABLE B6. Site Numbers available from Heritage Programs based on assigning their wetland
records to macrogroups and condition strata.
Macrogroup Stratum
Northern & Central Floodplain Forest
Northern & Central Swamp Forest
Appalachian, Interior Plateau & Prairie Fen
North American Boreal Bog & Fen
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow
Eastern North American Wet Shrub, Meadow & Marsh
Great Plains Freshwater Wet Meadow, Wet Prairie &
Marsh
Total
Condition Stratum
A
7
1
1
1
7
3
3
23
B
14
16
26
16
24
28
15
139
C
14
12
13
13
7
15
13
87
D
4
4
4
9
7
28
Total
39
33
44
30
38
55
38
111
Samples for the original design were chosen from the suite of existing sites randomly, using the
"Random Selection within Subsets" tool in the Hawths Tools extension for ArcGIS (Beyer 2004).
This tool picks a random subset of features from all features in a GIS shapefile using user-defined
strata. The selection of samples for one macrogroup is illustrated on Figure B4. The resulting
samples were reviewed to ensure adequate representation across ecoregions. Between the state
EO data and MiRAM data, we achieved over 95% of our target number of 280 points using known
sites. Two thirds of the data were to be collected in Michigan, one third in Indiana. The remaining
56
-------
5% of sites were attained by relying on the expertise of the field staff, who could suggest potential
sites where missing combinations of macrogroup and condition were needed.
FIGURE B4. Location of all sites and their EORANK for a macrogroup (Rich Fen)
based on state Heritage Databases, and categorized by whether or not they were
selected as part of the sample design.
MG061 Appalachian, Interior Plateau & Prairie Fen
IN
A
, Milwaukee
S«mU
Indianapolis
o
HoMfogtofl
25 50
100
• Miles
7^
..i.l.nilla
Proposed Assessment Sites
• Selected. A Rank
O Selected. B Rank
C Selected. C Rank
• Selected. D Rank
o not selected. A rank
° not selected. B rank
o not selected. C rank
° not selected. D rank
57
-------
In order to maximize the amount of real field time available and spend less time in transit, we
preferentially sampled sites located in somewhat close proximity to one another. This would speed
up field work considerably. However, in order to avoid spatial autocorrelation, we maintained a
minimum separation distance of 0.5 km for sites within the same macrogroup. This distance is
consistent with the surrounding landscape area used for the calculation of Landscape Context
metrics (Table B4). In practice, the vast majority of field sites within the same macrogroup were
located at least 1 km apart.
An overdraw of sites was made to account for the possible need to replace sites from this original
random draw.. The overdraw pool included one additionally randomly selected site per strata.
Because crews were state-based, separate overdraw pools were created for Indiana and Michigan.
Additional replacement sites could be selected later as necessary, by choosing randomly from the
remaining pool of known sites for that stratum.
We gave some consideration in our site selection to issues of public/private ownership. Our
random sample draw methodology described above did not explicitly take into account land
ownership or favor clustered samples. However, once a provisional selection of sites had been
made, we allowed field crews to swap sites based on accessibility. Publicly owned sites required
less preparation, because access permission is more readily obtained. Over 90% of the original
sites selected were part of the final sites sampled.
It was apparent that our overall pool of sites was relatively short on both A-ranked and D-ranked
sites, so crews were instructed to identify additional occurrences of such sites during the course of
their survey work, and determine if they met the needs of the sample design.
Caveats
Although designed to minimize bias, the methodology is not unbiased. There are geographic or
ownership biases that underlie sites catalogued by natural heritage programs. Consequently, we
cannot be certain that the suite of sites from which we chose our project sample represents the
true population distribution of reference sites. For those combinations of wetland type and
condition with fewer available Heritage sites (such as D-ranked sites), most or all existing sites were
included in the project sample. Thus, even as we attempted to maintain a geographic spread, any
spatial and ownership biases within the Heritage and MiRAM datasets were carried into our data
set. Likewise, supplementing the randomly chosen sample with additional sites identified based on
drive-by selection or local expert knowledge may have introduced additional biases into our
sampling methodology.
Despite these violations of truly random and unbiased sampling, we are confident the sample
design satisfied the goals of the project - to ensure that the full range of variation in wetland types
and conditions are sampled across the study area in an efficient manner. Given data limitations
present at this time, random selection methods were used to the greatest degree possible, and our
method allows inference with regard to the range of conditions of all major wetland types within
the project area.
STATISTICAL TESTS OF SAMPLING DESIGN
We evaluated our sampling design by examining the distribution of actual condition ratings (from
the field results of our ecological integrity assessments, as summarized in Section C) against the
predicted condition ratings from our condition stratum methods. Ideally, we would like to ask how
well these ratings might predict overall ecological integrity (based on landscape context, condition,
58
-------
and size), but comparisons with the overall IEI would be misleading because we used the same
Landscape Context ratings when calculating an overall IEI as we did for the condition stratum.
These measures are correlated by definition. So instead our main focus was on how well we
predicted on-site Condition and Vegetation. Our reference information was the Heritage EORANK,
which played no role in assigning our Ecological Integrity rating.
We chose a one-way analysis of variance (ANOVA) (appropriate to categorical rating data) to see
how well the condition stratum performed. We used the Kruskal-Wallis rank sum test, which is
the non-parametric analogue of a one-way ANOVA (when there is one nominal variable and one
measurement variable and the measurement variable does not meet the normality assumption of
an ANOVA). The Kruskal-Wallis test does not make assumptions about normality. Like most non-
parametric tests, it is performed on ranked data, so the measurement observations are converted
to their ranks in the overall data set: the smallest value gets a rank of 1, the next smallest gets a
rank of 2, and so forth. We applied the Pairwise Wilcoxon Rank Sum Tests to calculate pair-wise
comparisons between condition ratings with corrections for multiple testing, making adjustments
to p values when testing multiple comparisons. Finally, comparisons were scanned using Notched
Boxplots. If the two boxes' notches do not overlap, this is "strong evidence" that their medians
differ (Chambers et al, 1983, p. 62).
B.3 RESULTS
REFERENCE GRADIENT BY CLASSIFICATION STRATUM
Macrogroups and State Types
Our sampling design resulted in a range of wetland sites spread across the 7 macrogroups found in
the study area. We maintained a fairly consistent level of sampling across all macrogroups
(between 30 - 55 sites per macrogroup), spread across the ecoregions (Table B7, Fig. B5).
TABLE B7. List of Formations and Macrogroups found in the study area, and number of study sites
found for each macrogroup. Colloquial names for macrogroups are provided in square brackets.
NVC FORMATION
Macrogroup
Number
of Sites
SWAMP & FLOODED FOREST
Northern & Central Floodplain Forest (M029) [Floodplain Forest]
39
Northern & Central Swamp Forest (M030) [Swamp Forest]
33
FRESHWATER SHRUBLAND, WET MEADOW & MARSH
Eastern North American Wet Shrub, Meadow & Marsh (M069) [Wet Shrub, Meadow & Marsh]
55
Great Plains Wet Meadow, Wet Prairie & Marsh (M071) [Wet Prairie]
38
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow (M067) [Coastal Plain Pondshore]
38
BOG & FEN
Appalachian, Interior Plateau & Prairie Fen (M061) [Rich Fen]
44
North American Boreal Bog & Fen (M062) [Bog & Poor Fen]
30
Total 277
59
-------
FIGURE B5. Sites selected for the reference gradient of wetland types.
EIA Sites
Community Name
s Bog and Poor Fen
c Coastal Plain Pondshore
• Flood plain Forest
; Rich Fen
• Swamp Forest
C Wet Prairie
: Wet Shrub, Meadow, and Marsh
EPA Level III Ecoregions
Level III Name
Central Corn Belt Plains
Eastern Corn Belt Plains
Huron/Erie Lake Plains
Southern Michigan/Northern Indiana Drift Plains
Within each macrogroup, we also sampled all state natural community wetland types at least once
(Table B8).
60
-------
TABLE B8. State Natural Community Types by NVC Macrogroup.
Macrogroup
State & State Natural Community Type
Northern & Central Floodplain Forest (total)
IN (total)
Mesic Floodplain Forest
Wet Floodplain Forest
Wet-mesic Floodplain Forest
Ml (total)
Floodplain Forest
Northern & Central Swamp Forest (total)
IN (total)
BluegrassTill Plain Flatwoods
Boreal Flatwoods
Central Till Plain Flatwoods
Forested Fen
Sand Flatwoods
Swamp Forest
Ml (total)
Hardwood-Conifer Swamp
Rich Conifer Swamp
Rich Tamarack Swamp
Southern Hardwood Swamp
Wet-mesic Flatwoods
North American Boreal Bog & Fen (total)
IN (total)
Acid Bog
Ml (total)
Bog
Great Plains Freshwater Wet Meadow, Wet Prairie & Marsh (total)
IL (total)
Wet Prairie
IN (total)
Wet Prairie
Wet Sand Prairie
Wet-mesic Sand Prairie
Ml (total)
Interdunal Wetland
Lakeplain Wet Prairie
Lakeplain Wet-mesic Prairie
Wet Prairie
Number of
Occurrences
39
17
8
2
7
22
22
33
14
1
1
5
3
2
2
19
1
3
5
7
3
30
10
10
20
20
38
1
1
9
2
2
5
28
5
6
5
5
61
-------
Wet-mesic Prairie
Wet-mesic Sand Prairie
Appalachian, Interior Plateau & Prairie Fen (total)
IN (total)
Circumneutral Bog
Circumneutral Seep
Fen
Marl beach
Panne
Ml (total)
Prairie Fen
Eastern North American Wet Shrub, Meadow & Marsh (total)
IN (total)
Acid seep
Marsh
Sedge Meadow
Shrub Swamp
Ml (total)
Emergent Marsh
Great Lakes Marsh
Inland Salt Marsh
Inundated Shrub Swamp
Northern Wet Meadow
Southern Shrub-carr
Southern Wet Meadow
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow (total)
IN (total)
Muck Flat
Sand Flat
Ml (total)
Coastal Plain Marsh
Grand Total
4
3
44
15
3
3
7
1
1
29
29
55
24
1
11
4
8
31
6
5
3
1
1
3
12
38
12
9
3
26
26
111
Hydrogeomorphic Classification
For each wetland occurrence visited, we assigned the HGM class (Table B9). The kind and number
of HGM classes found within each macrogroup is reported in Table BIO. The most predominant
HGM type in the study was Depressional. Organic Flats are relatively rare in these ecoregions.
62
-------
TABLE B9. HGM type and number of occurrences for the project. "Primary only" refers to it
being listed as the only HGM class at a site.
HGM Class
Depressional
Lacustrine Fringe
Mineral Soil Flats
Organic Soil Flats
Riverine
Slope
Grand Total
Primary
Only
118
24
23
11
48
43
267
Primary +
Secondary
6
1
2
1
10
Types of Secondary Classes
Mineral Soil Flats (3), Organic
Soil Flats (1), Riverine (2)
Organic Soil Flats
Slope
Slope
TABLE BIO. The variation in HGM primary class within each NVC macrogroup.
Macrogroup: HGM type
Northern & Central Floodplain Forest
Mineral Soil Flats
Riverine
Slope
Northern & Central Swamp Forest
Depressional
Lacustrine Fringe
Mineral Soil Flats
Organic Soil Flats
Riverine
Slope
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow
Depressional
Lacustrine Fringe
Mineral Soil Flats
Organic Soil Flats
Eastern North American Wet Shrub, Meadow & Marsh
Depressional
Lacustrine Fringe
Mineral Soil Flats
Organic Soil Flats
Riverine
Slope
Great Plains Wet Meadow, Wet Prairie & Marsh
Depressional
Number of Sites
39
1
37
1
42
24
2
8
1
1
6
38
30
2
4
2
46
25
11
1
1
5
3
38
19
63
-------
Lacustrine Fringe
Mineral Soil Flats
Riverine
Slope
Appalachian & Interior Plateau Bog & Fen
Depressional
Lacustrine Fringe
Organic Soil Flats
Riverine
Slope
North American Boreal Bog & Fen
Depressional
Lacustrine Fringe
Organic Soil Flats
Grand Total
4
10
4
1
44
3
4
2
3
32
30
22
1
7
277
REFERENCE GRADIENT BY CONDITION STRATUM
Field crews collected data that permitted us to calculate the Ecological Integrity rating for each
wetland, as well as the on-side Condition and Vegetation scores (see Faber-Langendoen et al.
2011c). We can thus examine how well the overall predicted set of wetland conditions, based on
the "condition stratum" matched the final measured condition and vegetation ratings. Table Bll
shows the number of sites for each condition rating based on field measures, as well as the
predicted rating based on the individual or combination of factors used for the Condition stratum
Landscape Context alone, Heritage EO Rank alone, and various combinations of the two (see
Methods: Classification and Condition Strata above).
TABLE Bll. Actual and Predicted A- D Condition ratings based on various condition stratum
factors, alone or in combination. Not shown are the macrogroups.
Condition Stratum
Actual A - D condition ratings from field
Landscape Context
Heritage EO Rank
Condition Stratum 1 (Landscape Context X EO Rank -
rigorous A)
Condition Stratum 2 (Landscape Context X EO Rank -
moderate A)
A
127
57
55
23
66
B
126
122
78
139
135
C
23
73
73
87
59
D
1
25
20
28
17
Un-as-
signed
51
Grand
Total
277
277
277
277
277
64
-------
Our key question then is as follows: Did our condition stratum approach ensure that we sampled a
full range of wetland condition, based on both our original Condition Stratum approach 1 and
Condition Stratum 2? Our results are summarized in Table B12 and Figure B6 (see also Table Bll).
Looking first at Condition (Vegetation, Hydrology and Soils), we can see that on-site Condition is
best predicted by the combination of Landscape Context and Heritage EORANK, using either
approach 1 or 2, but our 2nd approach is an improvement (in which our requirements for an A
rating were not as rigorous) (Fig. B6). Thus, our first condition stratum approach underestimated
the number of A-rated sites compared to our second approach.
Our condition stratum 2 approach was also the best predictor of Vegetation ratings (Table B12).
Vegetation ratings are also less influenced by landscape context, and more informed by knowledge
of on-site condition, as provided through the Heritage EO ranks.
TABLE B12. F-values for one-way Kruskal-Wallis rank sum test on the various stratification
approaches: Landscape Context alone, Heritage EO Rank alone, Stratification Rank 1 (Landscape
Context X Heritage EORANK stringent A requirements) and Stratification Rank 2 (Landscape Context
X Heritage EORANK moderate A requirements). All F values have p <0.001, indicating all factors
successfully distinguish A/B from C from D. See also Figure B6.
Factors
Landscape Context
Heritage EO Rank
Condition Stratum 1
Condition Stratum 2
Condition Score
52.0
47.7
54.8
68.8
Vegetation Score
25.1
35.0
33.1
40.8
FIGURE B6. Comparison of predicted on-site Condition scores based on the
Condition stratum method 1 (Stratification Rank) and Condition stratum
method (Stratified Rank 2)
Stratification Rank
Stratified Rank 2
65
-------
B.4 DISCUSSION
CLASSIFICATION
In creating a sampling design for our reference gradient, our first concern was to sample the full
range of wetland types. As shown in Table 8, we were able to do this by linking the fine grained
state natural community types to the mid level NVC macrogroup. We also sampled the full range
of HGM types found in the region: we had a preponderance of Depressional types, and few Organic
Flats types. Thus, using a classification stratum based on NVC macrogroups ensured that our
sampling covered a wide range of wetland types. Because we could confidently assign sites to
macrogroups, we were rather successful in maintaining a balanced set of sites across macrogroups
(30 - 55 sites per macrogroup).
CONDITION
Our second concern for our reference gradient was to sample the full range of wetland conditions.
Our approach was to rely on a combination of factors to establish a condition stratum -, first the
readily observable landscape context factors available from imagery, and second, the expert based
field evaluations of on-site conditions (especially vegetation) recorded by state Natural Heritage
ecologists. We found that, by using both criteria, we were more successful at predicting the full
range of reference gradient conditions than either landscape context or on-site Heritage rank alone
(Table 12). Our field design was executed using condition stratum 1, which under predicted the
number of A-ranked sites we might encounter and over predicted the number of D-ranked sites.
Thus we sampled fewer D-ranked sites than were predicted by that approach. We recommend
using our revised condition stratum design for future studies.
Others have also found that landscape context metrics alone, based on remote sensing imagery,
have only limited value in predicting on-site condition (Mack 2006, 2007, Mita et al. 2007). This
suggests caution in using landscape alone as a predictor of individual site conditions, though it is
helpful for assessing overall watershed condition (e.g., Tiner 2004).
The stratification methods we use here bode well for identifying a reference gradient of wetlands
across many parts of the country. Heritage program data are widely available across the country.
Indeed, we have recently compiled all Heritage data into a master database, in which macrogroups
have been assigned to all state records. Currently available remote sensing imagery can be used to
calculate the landscape context metrics, and these can be combined with the condition stratum
method 2 to provide a robust prediction of on-site wetland condition. These data can be a primary
source of reference sites for studies needing a reference gradient for wetlands, grasslands, forests
and other types. In fact, EPA's National Wetland Condition Assessment is currently using these
data as part of their process to identify approximately 150 benchmark reference standard wetland
sites across the country to help inform their wetland condition assessment (Faber-Langendoen et
al. in prep, G. Serenbetz pers comm. 2011). Knowledge of these sites is becoming increasingly
important, given continuing levels of conversion or degradation of native ecosystems across many
parts of the country.
66
-------
SECTION C: TESTING THE ECOLOGICAL INTEGRITY
ASSESSMENT METHOD
C.I INTRODUCTION
There is a growing body of wetland assessment methods that provide standardized field sampling
and reporting methods for assessing wetland condition (e.g., Mack 2001, 2004, Collins et al. 2007,
Jacobs et al. 2010, see summary in Fennessy et al. 2007). Data on the ecological condition of
wetlands can be used for ambient monitoring of wetland status and trends, to prioritize sites for
conservation or restoration, guide mitigation applications at site and watershed scales (Faber-
Langendoen et al. 2008) and contribute to land use planning. Much has been done to develop and
test methods for assessing wetland condition, including both rapid and intensive methods (Mack
2001, 2004, 2006, Collins et al. 2007, Miller et al. 2006, 2007, Fennessy et al. 2007). But many
studies have been local in geographic scope, or restricted to a subset of wetlands in a region, or
have not provided a larger framework within which to assess changes in wetland condition. There
is a need to provide methods that assess condition across the full range of wetland types and
condition in a region and provide summary reports that make the results accessible to a wide range
of audiences.
Our purpose in this section is to apply our ecological integrity assessment method to the 277
wetland sites that span the reference gradient of wetland types and conditions across the study
areas (as identified in Section B above). We then completed a "post-hoc" analysis of the indicator
data collected across the sites to verify that the metrics and scores can discriminate among a range
of conditions, from "excellent" (minimally disturbed) to "poor" (degraded) wetlands. We also
created a scorecard and index of ecological integrity (IEI) for reporting on wetland integrity.
C.2 METHODS
Our site selection methods for the 277 sites are fully described in Section B (see Table B6, Fig. B5
for summaries). Here we describe the ecological integrity assessment field methods used to survey
each site.
LEVEL 2: RAPID FIELD ASSESSMENT METHODS
Overview of Field Methods
A field manual was developed to guide the use of field forms by the crews (Faber-Langendoen
2011). The general procedure for conducting a Level 2 assessment consisted of a series of steps
(adapted from Collins et al. 2006, Chapter 3):
PRELIMINARY SITE SELECTION (OFFICE)
Step 1: We reviewed the occurrence list for the field season, based on the sampling design
established for the project, including the primary sites and backup sites. Site selection was
determined by the wetland types and their conditions. The statistical design of the study was set
up to avoid sampling two wetlands in close proximity that were also of the same type (because the
67
-------
area sampled as part of the landscape context would overlap, making the samples within a type
non-independent).
Step 2: For each wetland occurrence at a site, we assembled background information about the
ownership, access, size, condition (=EORANK), spatial location (many sites had the spatial extent of
the wetland type mapped in GIS), and state wetland classification type. We also assembled aerial
photo imagery for the site. We made landowner contacts, as needed, before accessing the site.
Step 3: We reviewed the classification of the wetland, starting from the state Heritage Program
classification, which was nested within the U.S. National Vegetation Classification at the
macrogroup level (See Table B8 above). Additional classifications include NatureServe's Ecological
Systems (Table B3 above) and the Hydrogeomorphic (HGM) classification (Tables B9, BIO). Field
crews needed to identify the state natural community type and the HGM class.
State Natural Community Type:
a) Michigan. See Kost et al. (2007) and http://web4.msue.msu.edu/mnfi/
b) Indiana. See Jacquart et al. (2002) and
http://www.in.gov/dnr/naturepreserve/4743.htm
Descriptions for the macrogroups are being compiled and will become available on the
NatureServe website at: www.natureserve.org.and the USNVC partners website at:
Step 4: We verified the appropriate season and other timing aspects of field assessment for
sampling various wetland types.
SITE SELECTION REVIEW (OFFICE)
Step 5: We defined the assessment area (AA) as an area of given condition and wetland type (at
state natural community scale) and small enough in size to be observable in the course of a 2-4
hour visit (Rocchio 2007). Accordingly we define the AA as "the entire area, sub-area, or point of
an occurrence of a wetland type with a relatively homogeneous ecology and condition." Practically
speaking, this meant AAs had to be less than 20 ha (50 acres). For large wetland occurrences (> 20
ha), we determined which portion of the wetland could be visited. AAs could not always be
determined prior to the field visit, so adjustments were made in the field.
Although 20 hectares is too large to survey intensively, the crews made a judgment as to whether
the area they surveyed appeared typical of the entire AA or the polygon or EO within which the AA
occurred. The advantage of this approach is that a "polygon" or "wetland occurrence" focus is
maintained, rather than a "point-based" approach (See Fennessy et al. 2007, Faber-Langendoen et
al. 2011a). For many applications, the goals of the EIA are to determine the condition of an
extensive area of a wetland occurrence or polygon.
In some rapid assessments, the type and condition are ignored and the entire wetland is assessed
as part of the AA. In other assessments, detailed guidelines for establishing AAs are provided
(ORAM, CRAM). Our methodology follows the latter approach, consistent with Heritage
methodology, where an occurrence of a wetland type of conservation or management significance
is tracked based on its type, size and relatively uniform condition.
LANDSCAPE CONTEXT EVALUATION (OFFICE)
68
-------
Landscape Context metrics (LI) are extrapolated from remote sensing imagery. The goal is to
develop metrics that assess the landscape context (and thereby on-site conditions of an
ecosystem). Satellite imagery and aerial photos are the most common sources of information for
these assessments. Typically it is the stressors to the ecological integrity that are most observable
with these sources of information, so condition is generally inferred from stressors.
Step 6: Assess Landscape Context.
Primary Approach: Remote Sensing Landscape Metrics. NatureServe staff, using both satellite
imagery and aerial photography, established the buffer (200 m) around the Assessment Area (AA)
polygons, and, using the centroid of the polygon, established the circular areas that comprise the
"core" (100 ha) and "supporting landscape" contexts of the AA (1000 ha)10. We analyzed the
imagery to calculate the scores and ratings for the core and supporting landscapes and buffer
metrics for each occurrence at a site. The spatial boundaries of the landscapes and buffer and the
metric scores were moved into the database.
Secondary Approach: Landscape Condition Model. NatureServe has developed a Landscape
Condition Model (LCM, Comer and Hak 2009), similar to the Landscape Development Index used by
Mack (2006). The model provides a single stressor-based index that integrates the effect of
multiple landscape stressors on overall landscape condition. The algorithm for the model uses 30
m resolution pixels from various land use layers (roads, land cover, water diversions, groundwater
wells, dams, mines, etc.). These layers are the basis for various stressor-based metrics. The
metrics are weighted according to their perceived impact on ecological integrity, into a distance-
based, decay function to determine what effect these stressors have on ecological integrity. The
result is that each grid-cell (30 m) is assigned a "score". The product is either a watershed or
landscape map depicting areas according to their potential "integrity," or the condition of
individual polygons or patches can be characterized. The index is segmented into three or four
rank classes, from Excellent (minimally disturbed) (A) to Poor (degraded) (D) (See Appendix 2 for
more details).
ON SITE CONDITIONS (FIELD)
Step 7: For the field visit, standard field forms were used (Faber-Langendoen 2011). Further details
on the field methods are presented below.
7a. All sites were assessed using the rapid assessment (L2) method, including basic description
(vegetation and environmental characteristics), integrity metrics and stressor evaluation.
7b. One-third of the sites were further assessed using the intensive assessment (L3) method,
including a 0.1 ha plot. Plots focused on the vegetation, recording all species and their cover, and
recording stem diameters and density for all tree stems > 10 cm dbh, along with basic soil and
hydrology information to help characterize the wetland type.
DATA MANAGEMENT (POST FIELD)
10 In future versions of our protocol, we intend to use a "buffered polygon or point" approach, defining an
inner buffer of 100 m, the core landscape of 250 m, and the supporting landscape of 500 m. This provides a
more consistent landscape context assessment protocol for both buffer and landscape metrics.
69
-------
Step 8: All data were entered into an Ecology Observations Database (Access ©).
8a. Data clean-up was completed, including determination of appropriate plant species taxonomy
and ensuring the GPS coordinates accurately represent the location of the AA, etc.
8b. Data entry into NatureServe's Ecological Observations database was completed. Data in 2009
were stored separately for IN and Ml. In 2010, several changes were made to the protocol,
particularly for stressors, necessitating a slightly different design. The 2010 data from both states
were managed in a single database. The database provided a scorecard for each wetland and
export formats for analyses.
8c. QA/QC procedures were completed by the state program data entry staff and NatureServe
data management staff.
Step 9: Core data were uploaded into the state Heritage databases to upgrade site information,
classification, and EORANK.
Level 2 Field Protocols
A field crew (usually two people) typically conducts a rapid (L2) field assessment within two to
three hours, plus two hours preparation time assessing the imagery. Once the crew leaves the
field, the field forms are essentially complete, apart from data cleanup and QA/QC. Additional time
may be needed on plant species taxonomy issues, and to ensure that the GPS coordinates
accurately represent the location of the AA.
All field crews had at least one person who was trained in ecology, with sufficient botanical
expertise to recognize the major elements of the flora. Crews also had general experience with
hydrology and soils, sufficient for the rapid assessment methods. One-day field training exercises
were conducted in early May of each season in order to ensure consistent application of the field
protocols.
Upon arriving at a site, the crews were asked to validate the classification of the wetland
community to the state type, and thereby the NVC macrogroup (Table B8 above). If changes were
needed, crews documented these changes. Crews also assigned types to HGM class.
Typically, crews had pre-existing polygons to visit, based on Heritage element occurrence maps.
Crews checked the polygon boundaries of the map to verify or update the extent of the occurrence
as part of the field survey. Readily observable ecological criteria such as vegetation, soil, and
hydrological characteristics are used to define wetland boundaries, regardless of whether they
meet jurisdictional criteria for wetlands regulated under the Clean Water Act.
If the wetland was small enough to survey in its entirety, then the AA and the EO boundaries were
synonymous But if the EO was very large or if an EO had had two or more conditions present), then
the AA was restricted to a portion of the occurrence. Notes on the AA boundaries were made using
GPS and hand-drawn field notes on aerial photos and maps.
70
-------
Level 2 Metrics and Field Forms
Standard field forms were used by all field crews in both states (Faber-Langendoen et al. 2011d).
The list of metrics collected by the field crews is shown in Table C2. Some metrics vary by NVC
formation/macrogroup and HGM class, so crews used their assignment of the state to the state
type or HGM class to guide their use of metrics. Field forms are structured by major categories:
GENERAL DESCRIPTION (Site, Location, and Classification), DETAILED DESCRIPTION (optional
narrative form), VEGETATION PROFILE, ENVIRONMENTAL PROFILE, ECOLOGICAL INTEGRITY
METRICS (Vegetation, Hydrology, Soil, Size, Buffer), STRESSOR METRICS (Vegetation, Hydrology,
Soil, Buffer).
TABLE C2. List of Metrics collected for Level 2 Ecological Integrity Assessments. Protocols
for each metric are provided in Appendixes. Metrics in italics were later dropped after
statistical assessment (see Table CIO). A tidal wetland metric (Barriers to Landwater
Migration) is shown for completeness, but it is not applicable to the study area.
RANK FACTORS
LANDSCAPE CONTEXT
SIZE
CONDITION
MAJOR ECOLOGICAL
FACTORS
LANDSCAPE
BUFFER
SIZE
VEGETATION
HYDROLOGY
SOIL
METRICS
Connectivity
Land Use Index
Barriers to Landwater
Migration (tidal, not used)
Buffer Index
Relative Patch Size
Absolute Patch Size (ha)
Vegetation Structure
Organic Matter
Regeneration (woody)
Native Plant Species Cover
Invasive Plant Spp. - Cover
Increasers - Cover
Vegetation Composition
Water Source
Hydroperiod
Hydrologic Connectivity
Physical Patch Types
Soil Disturbance
Water Quality
71
-------
To survey the AA at the L2 level, field crews walked through and visually assessed the site, and
recorded the measures specified on the field form. 11
Level 2 Stressors Checklist
Stressor information can be a useful addition when evaluating the ecological integrity of an
occurrence (Kapos et al. 2002). Typically, they aid in further understanding the overall condition of
a wetland. Stressors were recorded by Major Ecological Factor (Buffer, Vegetation, Soils,
Hydrology). Stressors were listed if and only if they were observed or inferred to be occurring, and
were not included if they were projected to occur, but do not yet occur. Stressors were
characterized in terms of scope and severity. Scope is defined as the proportion of the occurrence
that can reasonably be expected to be affected by the stressor. Within that affected scope,
severity is the intensity of damage to the occurrence that can reasonably be expected from the
stressor. Stressor scope and severity are used to create field-based versions of stressor indices.
Individual Stressors as rated as Very High, High, Medium, Low, based on the combination of Scope
and Severity (see Appendix 3: Level 2 Protocols - Stressors Checklists).
Standardized stressor checklists have been developed for a variety of rapid assessment methods
(Collins et al. 2006, Faber-Langendoen et al. 2008, Collins and Fennessy 2010). Our list was
adapted from those sources (Faber-Langendoen 2011).
LEVEL 3: INTENSIVE FIELD ASSESSMENT METHODS
Integration with Level 2 Methods
Our Level 3 method builds on the Level 2 methods. Of the 277 sites that are visited, 88 sites also
had a Level 3 assessment completed. The level 3 data are restricted to a single representative area
- a 0.1 ha plot - within the AA, whereas the level 2 assessment is completed on the entire
Assessment Area. Notwithstanding these differences in scale, we compared the level 3
assessment data with the level 2 assessment data.
Vegetation Sampling Methods
We used a vegetation sampling protocol that provides information on basic vegetation composition
and structure and was suitable for extracting metrics for the Floristic Quality Index (FQI).
Plot Method
11 Although not used, here, it may be desirable in some circumstances to add a "quick" vegetation
plot to the level 2 evaluation. For example, a 50 m tape is laid, with 10 m flags on each outer end,
and vascular plant species presence and cover are recorded by strata by surveying the 10 m wide
area on each side of the tape). A few key structural attributes could also be collected (e.g., coarse
woody debris, number of stems > 30 cm dbh. The plot can be subjectively or objectively placed
within the AA to represent the typical heterogeneity within the AA. Such a plot would provide
valuable quantitative information on vegetation cover, species richness and abundance to
supplement the rapid assessment, but pushes the methodology towards a hybrid of Level 2 and
Level 3.
72
-------
A 20 m x 50 m releve plot was used to collect vegetation data. The method is based on the historic
Whittaker diversity plot method, modified by Peet et al. (1998), and used by other wetland
assessment methods (Mack 2004, 2007a,b, Rocchio (2007). The structure of the plot consists of
ten 100 m2 modules (a total of 1000 m2 or 0.1 hectare) which are typically arranged in a 20 m x 50
m array (Fig. Cl).
FIGURE C1. Releve Plot Method (from Peet et al.
1998). I = intensive modules 2, 3, 8,9.
10
*
9
I
I
8
I
I
f
7
v /
4
\
6
* f
5
-».
SO METERS
Plots were laid out using a 50 m measuring tape, extended as the centerline of the plot from an
origin. Starting at the origin (zero), a stake flag (or flagging tied to vegetation) was placed every 10
m. Red stake flags or flagging were placed at the 0, 40, and 50 m marks and green stake
flags/flagging at the 10, 20 and 30 m marks. This helped visualize the four "intensive modules"
which occur on either side of the centerline between the 10-30 m marks. Next, a 10 m rope was
extended perpendicular on either side of the centerline at each 10 m mark. Red or green flags
were placed at the end of the rope to mark the lateral boundaries of each module and the plot.
The plots were located subjectively by the field crews. Under typical conditions, the specific
location was chosen because it was judged to contain structure(s) and composition(s) typical of the
observed wetland, or to contain the most frequently occurring structure(s) and composition(s). If
the wetland had an irregular shape and 20 m by 50 m plot would not "fit" into the specified
wetland, the 2x5 array of modules were restructured to accommodate the shape of the wetland
or AA. For example, a 1 x 5 array of 100 m2 modules was used for narrow, linear areas. A 2 x 2
array of 100 m2 modules was used for small sites (Peet et. al. 1998; Mack 2004). Regardless of the
structure, a minimum of four intensive modules were always sampled.
Each module in the plot was numbered by standing at the 0 m mark facing the 50 m end (Figure 5).
The modules were assigned from 1-5 starting on the right side and modules 6-10 were assigned
using a similar method then from the 50 m mark. Intensive modules were typically number 2, 3, 8,
and 9. For those plots that did not use a 2x5 array of modules (e.g. 1x5 or 2x2), the module
numbers may be different (and were randomly chosen).
Floristic measurements include presence/absence, first made within the four core (or "intensive")
100 m2 modules (2, 3, 8, 9). Crews recorded all species in the first module, then added any new
species module by module for the remaining three modules. They evaluated percent canopy cover
of each species across all four modules (400 m2). They then searched the remaining six modules
("residuals"), adding any new species and cover values for those not found in the intensive
73
-------
modules (see Yorks and Dabydeen 1998 on the importance of this additional search area). They
also noted any species of interest found within the wetland AA but not in the plot. Thus the final
values obtained were a single species list with cover values across a 0.1 ha area. Tree species cover
was assessed separately for seedlings, saplings, and overstory. Overstory tree cover was assessed
across the entire 0.1 ha plot regardless of what modules they occurred in.
Cover was visually estimated at the level of the 100 m2 module (depth 1) using the following cover
classes (Peet et al. 1998): 1 = trace (one very small individual), 2 = 0.!- 95%.
Tree Structure
Live Stems
Level 3 vegetation sampling also included a stem profile. Information on size and number of tree
stems were collected by tallying tree stems (separately by species) to the nearest 10 cm intervals
from (1) -10 - 49 cm dbh, and recording the diameter of each stem to the nearest cm for stems
greater than 50 cm dbh. Stem information was collected on all stems > 10 cm dbh. Measuring the
diameter-breast-height (dbh) for each stem in the plot allows calculation of basal area, a widely
used measure of tree abundance, and density. Measures of dbh can also provide important
information on stand dynamics and structure not captured by cover estimates. This cut-off can be
lowered in environments where mature trees may often be much smaller or in cases where
information on the regeneration (sapling and/or seedling) layers is needed.
Standing Snags
All standing snags (dead standing tree boles >10 cm and at least 1.4 m tall) were recorded
Fallen Logs (coarse woody debris)
Fallen logs were defined as dead fallen tree trunks greater than 10 cm in diameter. Each fallen log
was recorded by diameter and length. The diameter was assigned using size class categories from
10 - 50 cm, and to nearest cm for over 50 cm. The length was assigned to nearest m for that part of
the stem that is within the plot and which exceeds 10 cm diameter.
Soil Sampling Methods
In addition, soil and substrate characteristics were collected at each site to aid in characterization
of the wetland at the site. A soil core or auger was used to estimate soil values in each of two
locations. A maximum depth of 50 cm was sufficient to record the following: Depth to Impervious
Layer, Depth to Saturated Soils, Depth to Water Table, Organic Soil (Sapric (muck), Hemic (mucky
peat), or Fibric (fibric peat)), Mineral Soil Texture, and any additional comments were recorded
about the soils (e.g., presence of marl layers, irregular depressions, mounds, etc.).
Sample Handling and Processing
Plant specimen data were handled in typical fashion (see Mack 2007b for details). Standard state
nomenclature was used, and then standardized to PLANTS / NatureServe taxonomy for vascular
plant, nonvascular plant, and lichen names as accepted by NatureServe's standard references,
which represent the consensus standards for researchers working in a given geographic area (see
Kartesz 1999).
74
-------
We coordinated this standard nomenclature with the standard state nomenclature used in
Michigan and Indiana.
DATA MANAGEMENT
Level 2 Data Management
All data were entered and managed in an Ecological Observations Database that was specifically
designed for the project, yet structured as generically as possible to provide an ongoing database
tool for other ecological integrity assessment projects. The database is structured to match field
data protocols: General Site Description, Level 2 metrics, Level 2 stressor checklists, and Level 3
metrics, including vegetation plot data. Data in 2009 were stored separately for IN and Ml. In 2010,
several changes were made to the protocol, particularly for stressors checklists, necessitating a
slightly different design. The 2010 data from both states were managed in a single database. An
Index of Ecological Integrity (IEI), including a scorecard, was used within the database to summarize
all metric ratings for L2 assessments (see Index of Ecological Integrity section below). Data are
available from NatureServe and from the Natural Heritage Programs upon request.
LEVEL 3 Vegetation Data management
Vegetation data were entered into a Microsoft Access™ database, where cover class data were
transformed into cover values (the midpoint of each cover class). Mean cover for each species
were averaged across the intensive modules and used in data analysis. For those species only
occurring in the residual plots, the cover value for the residual plots was used for analysis. To
eliminate spelling errors, a drop-down list was used for species entry. Unknown or ambiguous
species (e.g. Carex sp.) were recorded but not included in data analysis. Data entry was reviewed
by an independent observer for quality control.
The Michigan Floristic Quality Assessment (FQA) database (Herman et al. 2001) was used to
populate life history traits, wetland indicator status, and C-values in the data reduction spreadsheet
for each species in the plot. Species nomenclature follows USDA PLANTS Database
http://plants.usda.gov/) as of January 2009. Since many practitioners in Michigan use the Michigan
Flora by Voss (1972, 1985, 1996), and in Indiana use a variety of floras as a field key and
nomenclature reference, these names were cross-referenced to the PLANTS names in the FQA
database.
75
-------
INDEX OF ECOLOGICAL INTEGRITY
IEI and Scorecard
As described above, our methodology included a prototype Index of Ecological Integrity (IEI) for
Level 2 assessments. Our selection of metrics for the IEI was based on reviewing tests of these
metrics from the literature (e.g. for ORAM see Mack 2006, 2007a, for CRAM see Sutula et al. 2006,
Stein et al. 2009). But, we also conducted our own statistical tests to validate the method, and
adjusted the use of metrics before generating a final scorecard (see below). The IEI was generated
from the Ecological Observations database after all metrics had been scored in the field. The IEI is
presented in a summary scorecard that shows the scoring of all metrics, Major Ecological Factors
(MEFs) and primary Rank Factors. Most metrics within each MEF typically received a weight of 1.0.
Some were weighted as 0.5 if they were known to be partially redundant with other metrics (e.g.,
native species cover and invasive species cover metrics) or not as responsive as other metrics. Each
MEF received a weight of 1, except for Soils, which received 0.5. Finally, the overall Rank Factors
were weighted as follows: Landscape Context - 0.25, Size - 0.15, and Condition - 0.60, based on best
professional judgment of how these factors contribute to the overall condition, resistance and
resilience of a wetland, as reported in Faber-Langendoen et al. (2011a).
We exported all of the field values from the Databases into formats suitable for analyses. We
evaluated the data at multiple indicator levels: metrics, MEFs, and Rank Factors. Each site was a
row in a spreadsheet, and each column contained classification and other attribute information as
well as the various levels of indicators (over index scores, rank factor and major ecological factor
scores, as well as individual metric or indicator scores, and human stressor index scores).
STATISTICAL SCREENING OF METRICS, ATTRIBUTES AND INDEX OF ECOLOGICAL INTEGRITY
A central question is whether our assessment methods properly grade the wetlands from Excellent
to Poor across different wetland types. To evaluate the IEI, we conducted both statistical tests and
heuristic evaluations based on previous Heritage ranks.
Typically we relied on overall rank or categorical values of individual metrics, and used non-
parametric analyses. This is because our field observations typically involved assigning a rating of
A, B, C, and sometimes D or E to a metric, rather than a numeric score. In addition, we assigned
the overall IEI ratings using two variants: A, B, C, CD, and D (5 ratings), or to treat the CD as a subset
of C (i.e., 4 ranks = A, B, C (including CD), and D). But for some analyses we chose to treat CD as
part of D (4 ranks = A, B, C+ and CD/D), given the paucity of fully D-ranked sites. Could this be said
more clearly as "using two variants: A, B, C, CD, and D (5 ratings). Although sometimes CD ranks
were lumped with C ranks or D ranks due to the paucity of sites ranked CD or D."
Human Stressor Index
We developed a Human Stressor Index (HSI), following Rocchio (2007), based on a rollup of the
individual stressors impact scores reported by field crews for buffer, soils and hydrology (see Table
C3). The HSI is = Soils Rating + Hydrology Rating + Buffer Rating / 3. This stressor-based index is
primarily independent of the ecological integrity metrics, though some integrity metrics partly
consider stressors. To ensure complete independence in the analysis, we also compared the HSI
against Vegetation metrics alone.
76
-------
TABLE C3. Rollup procedure for creating an overall stressor rating for Buffer, Soils,
and Hydrology, which together are used to create a Human Stressor Index (see text).
Impact Values of Stressor Categories
1 or more Very High Stressors, OR
2 or more High, OR
1 High + 2 or more Medium
1 High Stressor, OR
3 or more Medium, OR
2 Medium + 2 or more Low, OR
1 Medium + 3 or more Low
1 Medium Stressor + 5 or more Low
Or 8 or more Low
1 Medium + 1-4 Low
1- 7 Low Stressors
0 Stressors
OVERALL STRESSOR RATING
(points)
Very High (1)
High (2)
Medium (3)
Low (4)
Absent (5)
Screening Metrics
We screened and scored the L2 and L3 metrics using comparable methods for other Indices of
Biotic Integrity (IBIs) (Barbour et al. 1996, Blocksom et al. 2002, Klemm et al. 2003, Jacobs et al.
2010). We examined discriminatory power or responsiveness, and redundancy.
Discriminatory power is the ability of a metric to distinguish high stress from low stress sites, based
on the Human Stressor Index (HSI), which integrates stressor scores for hydrology, soils, and buffer
into an overall score, which is then converted to a stressor rating (High, Medium, Low, Absent). We
compared how well various components of the EIA were able to distinguish these sites.
We evaluated metrics by examining their distributions using box-and-whisker plots and one-way
analysis of variance (ANOVA) (appropriate to categorical data for L2) to see if any of the metrics
had significantly different mean values among the four levels of Stressors. We used the Kruskal-
Wallis rank sum test, which is the non-parametric analogue of a one-way ANOVA (when there is
one nominal variable and one measurement variable and the measurement variable does not meet
the normality assumption of an ANOVA). The Kruskal-Wallis test does not make assumptions
about normality. Like most non-parametric tests, it is performed on ranked data, so the
measurement observations are converted to their ranks in the overall data set: the smallest value
gets a rank of 1, the next smallest gets a rank of 2, and so on. We applied the Pairwise Wilcoxon
Rank Sum Tests to calculate pairwise comparisons between group levels with corrections for
multiple testing, making adjustments to p values when testing multiple comparisons. Finally,
comparisons were scanned using Notched Boxplots. If the two boxes' notches do not overlap, this
is "strong evidence" that their medians differ (Chambers et al, 1983, p. 62). Variables with non-
significant F-values from the ANOVA were considered non-responsive and candidates for removal.
Redundancy was evaluated with the aim of minimizing metrics that were redundant. A key concern
for Level 2 evaluations is to keep the assessment time as efficient as possible. Metrics that
duplicate other metrics are candidates for elimination. We evaluated metrics and Major Ecological
77
-------
Factors (by calculating a Spearman rank correlation coefficient matrix among all metrics and MEFs,
and reviewing scatterplots (in part to be aware of non-linear patterns). Metrics correlated with an
r [s] > 0.8 were considered largely redundant (Jacobs et al. 2010), metrics correlated with an r [s]
0.6- 0.79 were considered partially redundant.
When redundant attributes are identified, the one with the strongest correlation to human
disturbance and most effective discriminatory power were considered most valuable to retain. If
redundant attributes (e.g. woody regeneration and % non-native species) provided unique
ecological information (level of abundance of woody saplings and seedlings vs. change in
abundance of non-native species) they were retained.
LEVEL 2 AND LEVEL 3 METRICS (COEFFICIENT OF CONSERVATISM)
Mean CC
We used the simple form of the coefficient of conservatism (CC): presence/absence, based on
testing by Rocchio (2007). We assessed the relationships between mean CC scores calculated from
Level 3 plot data (88 sites) and that of individual L2 metrics, overall vegetation scores, condition,
and overall integrity for those some 88 sites. We also evaluated whether the mean CC should be
used as a supplement to the more rapid L2 metrics, as one way to add a robust metric for rapid
assessments. That is, we assessed whether it was redundant with L2 metrics, and whether it added
value as a more quantitative measure.
We did not pursue testing or development of a Vegetation Index of Biotic Integrity (VIBI). VIBIs are
available for forest, shrub and herb wetlands in Ohio in some of the same ecoregions and
macrogroup as sampled here (Mack 2004, 2007a) and initial testing could start by using those VIBIs.
Vegetation Structure - Floodplain and Swamp Forests
A number of Level 3 metrics are available to us from forested sites, but we are not able to evaluate
them at this time, nor are they generally applicable, so they are not scored here. Typical Level 3
metrics could include:
• Structural stage (assessment of the old growth status of forested wetlands) (tree stems >
30cm dbh, > 50 cm dbh)
• Overstory tree basal area
• Overstory tree density
• Sapling density
• Volume of coarse woody debris
TESTING AND APPLYING THE REVISED El A METHOD
Based on revisions to the EIA method from statistical tests, we then assessed how responsive the
revised attributes and indices were to the Human Stressor Index. We summarized the IEI scores for
our 277 sites.
78
-------
C.3 RESULTS
We obtained data from 277 sites, of which all contained L2 data and 88 contained L3 data. We
focus on the overall patterns of condition and ecological integrity ratings, across all macrogroups.
STATISTICAL SCREENING -REDUNDANCY
We screened 23 primary metrics, including those that were components of a metric index (e.g., the
Connectivity metric index contains two primary (or sub-) metrics - core connectivity and supporting
connectivity). We also screened the metric indices (connectivity, land use, buffer), for a total of 26
metrics.
Landscape Context Metrics
The submetrics within both the Land Use Index and the Connectivity Index (core versus supporting)
were moderately correlated (and thus only partially redundant) (r = 0.70 for Connectivity, r = 0.71
for Land Use), supporting the use of these two scales. However, the indices themselves, were
strongly correlated (and thus considered redundant) (r = 0.89). This is not too surprising given that
both metrics rely on similar information from the imager - Connectivity assesses the percent
natural ecosystems in the surrounding landscape, and Land Use Index further assesses the intensity
of land uses within the cultural part of the landscape. Connectivity is straightforward and simple to
measure and would appear to be the preferred metric, at least in this region of the U.S. However,
the range of land uses encountered in our study (typically rural to wild landscapes) may not have
been widely representative, and we suggest that land use be retained until testing across a greater
variety of landscapes is completed.
The Buffer Index was only moderately correlated (and thus only partially redundant) with either
Connectivity (r = 0.69) or Land Use (r = 0.68). Given the distinctive evaluation needs and
characteristics of the buffer (and its extent, width and condition being assessed), we suggest that
buffer should be treated as its own MEF, separate from other Landscape Features.
Size
Relative Size (percentage reduction in wetland size from draining, filling etc. compared to its
natural size) and Absolute size were minimally correlated (r = 0.57), suggesting they were only
minimally redundant, and provide independent information on Size. Absolute Size was typically not
completed by field crews (147 sites not rated), as further research would have been needed to
determine optimal size rating scales for wetlands in these ecoregions.
Condition
Within Condition, the MEFs are only minimally correlated (r = 0.31-0.50), indicating that they are
minimally redundant, and thus worth assessing separately (Table C4). Of the 3 MEFs, Soils showed
the lowest correlation (r =0.50) with Condition, and Vegetation the highest (r = 0.90).
79
-------
TABLE C4. Correlation (spearman's rank) between the three Major Ecological Factors of Condition and
Mean Coefficient of Conservatism (CC).
MeanCC Condition Vegetation Hydrology Soils
Mean CC
Condition
Vegetation
Hydrology
Soils
1.00
0.55
0.61
0.31
0.25
1.00
0.90
0.78
0.50
1.00
0.50
0.31
1.00
0.36 1.00
Composition was the single metric strongly correlated with the overall Vegetation rating (r = 0.88)
and with Condition (r = 0.83), suggesting that field evaluation of this metric provided a strong
indication of the overall Condition of the site. Mean CC was moderately strongly correlated with
Vegetation (r = 0.61), but showed little to no correlation with Hydrology or Soils.
VEGETATION
Within Vegetation, only one pair of metrics was strongly correlated (and thus redundant); percent
natives and percent invasive exotics (r = 0.82) (Table C5). This is not surprising, given they are
essentially two sides of the same coin, the one from the perspective on condition (percent natives),
the other that of the stressor (invasives). Overall Composition was moderately correlated with
Structure, Natives, and Invasives (r = 0.61 - 0.64). Organic Matter had little correlation with any
other metric within Condition (r = 0.04 - 0. 37). Increasers also had low correlations (r =-0.10 to
0.46). Thus for future assessments, we suggest that Organic Matter and Increasers could be
dropped (see below for summary). Although either the Natives or Invasives metrics could be
dropped, given their correlation, they do provide important perspectives on the condition of
vegetation (e.g., some exotics present may not be invasive).
80
-------
TABLE C5. Correlations among vegetation metrics. Correlations above 0.60 are highlighted
Structure
Organic
Increaser
Native
Invasive
Regeneration
Composition
Mean CC
Structure
1.00
0.43
0.32
0.43
0.35
0.40
0.64
0.40
Organic
1.00
0.04
0.09
0.07
0.37
0.32
0.26
Increaser
1.00
0.33
0.30
0.16
0.46
0.38
Native
1.00
0.82
0.31
0.63
0.38
Compo-
Invasives Regeneration
sition
1.00
0.17 1.00
0.61 0.47 1.00
0.42 0.53 0.50
Mean
CC
1.0
Mean CC showed minimal correlations with other metrics, the highest being with the Composition
(r = 0.50) and regeneration (r = 0.53) metrics. As noted above, it did show a moderate correlation
with overall Vegetation scores, suggesting that it more strongly reflects the combination of
vegetation metrics, rather than any one metric.
HYDROLOGY
The three metrics were only minimally correlated (r = 0.25-0.46) (and thus minimally redundant).
The three metrics were minimally correlated with overall Condition (r = 0.56 - 0.67), though
collectively the correlation between Hydrology and overall Condition was fairly strong (r = 0.78)
(Table C4 above).
SOILS
The three metrics were uncorrelated with each other (r = 0.12 - 0.15), and only minimally
correlated with overall Condition (r = 0.25 - 45), with patch diversity having the strongest
correlation.
STATISTICAL SCREENING - DISCRIMINATORY POWER
Screening Rank Factors and Attributes
We compared how well various components of the EIA were able to discriminate between sites
with different levels of stressors, based on the Human Stressor Index (HSI), from High (including
both High and Very High) to Absent (Table 6). Landscape Context, Size, and Condition
discriminated among each of the stressor levels Within Condition, Soils showed only weak
discrimination, separating highly stressed sites from others. Hydrology and Vegetation were very
effective (Table C6, Figs. C6 and C7).
81
-------
TABLE C6. F-values for one-way Kruskal-Wallis rank sum test on the four stressor levels (High,
Medium, Low, Absent) for MEFs and metrics. Significance of F values are shown, and all
pairwise comparisons are significant at p < 0.001 (<«), p<0.01 («) or p <0.05 (<). Non
adjacent means are equal if underlined. Discriminatory Power ranges from High (all Stressor
Levels distinguished) to Poor (no Stressor levels distinguished).
EIA Score
ElARank
Landscape
Size
Condition
Hydrology
Soil
Vegetation
Mean CC
F-value
124.7***
128.7***
89.5***
59.4***
109.6***
129.2***
33.6***
59.6***
10.8*
Pairwise comparison
A>»L>»M>»H
A>»L>»M>»H
A»L>»M»H
A>»L>»M>»H
A»L>»M>»H
A»L>»M>»H
A=L=M>H
A=L>»M»H
A=L=M=H
Discriminatory
Power (H,M,L,
P)
H
H
H
H
H
H
L
M
P
FIGURE C6. Notched Boxplot of Rank Factor Ratings (x axis) in relation to Human Stressor Index
(yaxis).
ABSENT LOW MEDIUM HIGH
ABSENT LOW MEDIUM HIGH
Landscape Context
Size
Condition
82
-------
FIGURE C7. Notched Boxplot of MEF Ratings (x axis) within Condition in relation to Human Stressor
Index (y-axis).
ABSENT LOW
MEDLJM HK3H
ABSENT LOW
MEDIUM HK3H
ABSENT LOW
MEDIUM HIGH
Hydrology
Soils
Vegetation
Screening Metrics
We assessed each metric in turn, to determine its discriminatory power with respect to response to
stressors. We used the following ratings:
High (H) = F value p <0.001, and at least 4 stressor levels distinguished.
Moderate (M) = F value p <0.001 and at least 3 stressor levels distinguished
Low (L) = F value p < 0.001 and at least 2 stressor levels distinguished.
Poor (P) = F value p < 0.05 and 1-2 stressor levels distinguished.
All metrics, MEFs and Rank Factors had statistically significant contributions in discriminating among
sites categorized by stress. Our concern here is to ascertain those that make the most ecologically
important contribution to discriminating among these sites, recognizing that responses to abiotic
stressors are not necessarily the entire picture.
Landscape Context Metrics
All three landscape context metrics had moderate to high discriminatory power, leading to the overall
high discriminatory power of the rank factor (Table C7). Because all metrics are based on aerial photo
interpretation, all metrics were scored for all sites.
83
-------
TABLE C7. Discriminatory power of Landscape Context Metrics in response to the Human Stressor
Index. F-values for one-way Kruskal-Wallis rank sum test on the four stressor levels of the Human
Stressor Index (High, Medium, Low, Absent) for metrics. Significance of F values are shown, and all
pairwise comparisons are significant at p < 0.001 (<«), p<0.01 («) or p <0.05 (<). Non adjacent
means are significantly different unless underlined.
F-value
Pairwise comparison
Discriminatory
Power (H,M,L, P)
LANDSCAPE
Connectivity
Land Use Index
Buffer Index
89.5***
62.0***
74.0***
80.1***
A»L>»M»H
A»L>»M»H
A»L>»M»H
A>»L>»M=H
H
H
H
M
Size Metrics
The two size metrics showed moderate discriminatory power, and together provided an overall high
discriminatory power for the rank factor and attribute (Table C8). But field crews also had difficulty
scoring absolute size; 147 sites were not given ratings. 35 sites were not given Relative Size ratings.
TABLE C8. Discriminatory power of Size Metrics based on Human Stressor Index. See Table C7 for
details.
F-value
Pairwise comparison
Discriminatory
Power (H,M,L, P)
SIZE
Absolute Size
Relative Size
59.4***
33.77***
54.4***
A>»L>»M>»H
A=L»M>H
A=L»M»H
H
M
M
Condition Metrics
Condition metrics were organized by Major Ecological Factors of Vegetation, Hydrology, and Soils.
Vegetation overall had moderate discriminatory power to the abiotic stressor index; only the
Composition metric had a moderate discriminatory response, the rest had low (Table C9). Regeneration
was left blank or assigned a Non Applicable rating at 191 sites. Hydrology overall had high
discriminatory power, with all metrics having moderate discriminatory power. All metrics were scored
at all sites. Soils overall had low discriminatory power; patch diversity had a moderate rating, the others
poor, including Water quality, which was left blank or assigned a Not Applicable rating at 56 sites.
Given that the stressor index is based on abiotic stressors, it is perhaps not surprising that hydrology had
the best discriminatory power. Vegetation metrics appeared to be only partly responding to the
84
-------
assessed abiotic stressors. However, soil metrics appeared to be of little value in assessing response to
stressors.
TABLE C9. Evaluation of Discriminatory Power of Condition Metrics based on Human Stressor Index. See
Table C7 for details.
F-value
Pairwise comparison Discriminatory Power
(H,M,L, P)
CONDITION
Vegetation
Structure
Regeneration
Organic Matter
Natives
Invasives
Increasers
Composition
Hydrology
Water Source
Hydro. Connectivity
Hydroperiod
Soil
Patch Diversity
Soil Surface
Water Quality
Mean CC
109.6***
59.6***
37.1***
20.3***
26.6***
25.2***
23.3***
15.0**
59.9***
129.2***
59.5***
83.1***
98.6***
33.6***
30.0***
10.2*
8.6*
10.8*
A»L>»M>»H
A=L>»M»H
A=L»M=H
A=L»M=H
A=L>M=H
A=L>M=H
A=L>M=H
A=L=M>H
A=L>»M>»H
A»L>»M>»H
A»L>»M=H
A=L>»M»H
A=L>»M>»H
A=L=M>H
A=L>M>M
A=L=M=H
A=L=M=H
A=L=M=H
H
M
L
L
L
L
L
L
M
H
M
M
M
L
M
P
P
P
FINAL SELECTION OF METRICS
Vegetation metrics and MEFs were reviewed once more to ensure correlations were not based on
outliers and that each was ecologically meaningful. We compared pairs of metrics that were redundant,
and those that had the strongest correlation to human disturbance and the most effective
discriminatory power, based on the 2nd part of our screening, were considered most valuable to retain.
Only two pairs of metrics were flagged as redundant. The first pair, connectivity and land use, were both
highly correlated. Both also showed moderate discriminatory power. The 2nd pair, native and invasives
were both highly correlated, but only of lower discriminatory power with respect to abiotic stressors.
For now we reduced the weight of each metric to 0.5, and in the future suggest they could be combined
into a Native-lnvasives Species Index, as there is value in knowing both the total native species cover
and the cover of invasives.
In terms of discriminatory power, we proceeded cautiously, because discrimination was based on abiotic
stressor gradient, and not all integrity changes are due to abiotic stressors. If some redundant or low
85
-------
discriminating metrics (e.g. woody Regeneration and Invasive Species) provided unique ecological
information (level of abundance of woody saplings and seedlings vs. change in abundance of non-native
species) they were retained.
Based on this review, we simplified our set of metrics for wetland ElAs, reducing the total number from
18 to 15, with 1 optional metric, as shown in Table CIO. For Landscape Context, we suggest that Land
the Use Index is optional, but before we can recommend dropping it, we would like to test it across
more land use types in different regions of the country. Size remains unchanged. For Vegetation, we
removed the Organic Matter and Increasers metrics. Hydrology remains unchanged. For Soils, we
removed the Water Quality metric.
TABLE CIO. Revised set of metrics for wetland Ecological Integrity Assessments. Rd = Redundancy. DP =
Discriminatory Power.
RANK FACTORS
LANDSCAPE
CONTEXT
SIZE
CONDITION
MAJOR
ECOLOGICAL
FACTORS
LANDSCAPE
BUFFER
SIZE
VEGETATION
HYDROLOGY
SOIL
METRICS (original)
Connectivity
Land Use Index
Buffer Index
Relative Patch Size (ha)
Absolute Patch Size (ha)
Vegetation Structure
Organic Matter
Regeneration (woody)
Native Plant Species
Cover
Invasive Plant Spp. -
Cover
Increasers - Cover
Vegetation Composition
Water Source
Hydroperiod
Hydrologic Connectivity
Physical Patch Types
Soil Disturbance
Water Quality
Rd
Y
Y
Y
Y
DP
H
H
M
M
M
L
L
L
L
L
L
M
M
M
M
M
P
P
METRICS (revised)
Connectivity
Land Use Index (optional)
Buffer Index
Relative Patch Size (ha)
Absolute Patch Size (ha)
Vegetation Structure
Removed
Regeneration (woody)
Native Plant Species Cover
[weight 0.5]
Invasive Plant Spp. - Cover
[weight 0.5]
Removed
Vegetation Composition
Water Source
Hydroperiod
Hydrologic Connectivity
Physical Patch Types
Soil Disturbance
Removed
86
-------
APPLYING THE FINAL MODEL
Comparing Integrity Scores to the Stressor Index
We examined the relationship between the ecological integrity scores and the stressors acting on the
site, both on site and observed in the buffer using both the revised model, using the Human Stressor
Index to examine relationships.
We found a moderately strong correlation between the HSI and the IEI (r = 0.66), suggesting that our
measures of ecological integrity do moderately well at showing responses to human stressors (Table
Cll). The condition rank factor showed the greatest response, and within Condition, the Hydrology
attribute showed the strongest response. This may reflect how stressors to hydrology are the most
readily observable (e.g., ditches, dikes), as are their effects on hydrologic condition.
Neither Mean CC nor Heritage EO Rank showed a particularly strong relation to the HSI,
All individual metrics showed minimal correlations with the HSI (all r < 0.60), suggesting that multiple
metrics are needed to assess the combinations of stressors acting on a site.
TABLE Cll. Correlation of ecological integrity rank factors and major ecological factors to the Human
Stressor Index based on the final set of metrics for the EIA model.
RANK FACTOR
Major Ecological Factor
LANDSCAPE
Landscape
Buffer Index
SIZE
CONDITION
Vegetation
Hydrology
Soil
Index of Ecological Integrity
Mean_C
EO_RANK
HSI Rating
0.57
0.50
0.54
0.46
0.61
0.45
0.67
0.32
0.66
0.30
0.34
Comparing Landscape Context and On-Site Condition
We assessed the degree to which on-site Condition ratings were correlated to surrounding LI Buffer and
Landscape metrics. The overall Landscape Context Rank Factor had a modest correlation (r = 0.47) with
on-site Condition, and among the individual metrics, the Buffer Index was the strongest (r=0.48) (Table
C12). Within overall Condition, Hydrologic Condition was most strongly correlated with Landscape
Context and its metrics (r = 0.52 - 0.54). In addition, the Human Stressor Index had a modest correlation
with Landscape Context metrics (0.52-0.57).
87
-------
TABLE C12. Landscape Context metrics versus Condition and Stressor Index ratings. The
Landscape Context metrics are based on Level 1 metrics; the Condition and Stressor Index on
Level 2 metrics.
CONDITION
Vegetation
Hydrology
Soils
Mean_CC
HSI
LANDSCAPE
CONTEXT
0.47
0.34
0.54
0.18
0.40
0.57
Connectivity
0.37
0.26
0.46
0.13
0.39
0.47
Metrics
Land Use
0.41
0.28
0.48
0.15
0.34
0.52
Buffer Index
0.48
0.36
0.52
0.21
0.37
0.54
Level 3 (mean cc) and Level 2 metrics
Our evaluation of the responsiveness of the mean CC values (from 88 sites) showed that it did not
provide a readily interpretable part of on-site condition or overall Ecological Integrity. It added little to
what we had learned from the Level 2 metrics (scored across all 277 sites).
Level 1 Predictions of Level 2 Ratings
We tested how well Level 1 ratings predict Level 2 ratings at individual sites. We compared two
approaches that used the Landscape Context metrics around a site to predict on-site condition. The first
approach is based on the general Landscape Condition Model of Comer and Hak (2009) (Fig. 8a), as
summarized in Appendix A; the 2nd is based on our Landscape Context metrics of the EIA model used
here. Both models do a reasonable job of separating out differences in Condition, but the general
Landscape Condition Model is less successful than the Landscape Context Ratings, including being able
to recognize only 3 levels of condition, rather than the 4 levels of the latter (Figs C8a, C8b). However,
the models are more successful in predicting overall IEI scores (not shown), because landscape context
and size, as well as on-site condition are relevant to an IEI, and level 1 models do well at assessing these
aspects of integrity. Although not attempted here, it may also be possible to re-calibrate the Level 1
metrics used for Level 1 assessments based on these Level 2 scores in order to improve the prediction of
Level 2 Condition from Level 1 metrics.
88
-------
FIGURE 8. Correlation between two kinds of Level 1 EIA (remote sensing) models and Level 2 (field-
based) assessments of On-site Condition (with scale of A (5.0) to D (D= 1.25). nd the Level 1 EIA Method
has a combination of stressor and integrity based metrics.
FIGURE 8a. The Landscape Condition Model rating (based on stressor-based metrics) for the combined
area of wetland and core landscape (1 km buffer around wetland). Kruskal-Wallis F = 32.6, p < 0.001.
VL= L< M< H =VH.
Very High High
Landscape Condition Model Rating
FIGURE 8b. Level 1 EIA Landscape Context ratings (based on three landscape context metrics of Table
CIO). Kruskal-Wallis F = 52.0, p < 0.001. A>B» C»D.
Landscape Context Rating
COMPARISON WITH NATURAL HERITAGE RANKS
Our presumption in designing the study was that previous evaluations of wetland condition by Heritage
ecologists were somewhat inconsistent, but in line with the ecological integrity approach. We would
like to determine more carefully just how the two are related. But it would be hard to use the provided
EORANKs. This is because previous Heritage evaluations may have occurred anytime in the last 2 to 20
years, so the field conditions of the wetlands may have changed from the previous visit. Instead, the
Michigan Heritage staff ecologists decided to revise their EORANKS based on the visits over the last two
years using their standard methods. There is some potential circularity here because they participated
89
-------
on the field teams completing the ElAs. But they had no knowledge of the preliminary or final IEI ratings
assigned here. Thus we felt the comparison would still be helpful. We compared the findings of our IEI
with those of the original ranks assigned by Heritage programs using Spearman's rank correlation
coefficients (r[s]). Spearman's rank is used because the ranks are ordinal data.
Michigan re-ranked 125 sites based on the 2009-2010 data. We first note that, for Michigan, the
correlation between the Heritage ranks pre-2009 with those of 2009-2010 was r = 0.48 (p <001).
Despite the statistically significant correlation, it was not a high correlation. Thus the new Ml ranks
changed considerably from the previous ratings.
Overall correlation between IEI numerical score and the new Ml EORANK was r = 0.62, whereas
correlations with the IEI categorical rating and Ml EORANK was 0.53. Ml Condition rating correlated
most strongly with the EIA Vegetation rating (r = 0.70), and, more surprising, was correlated r = 0.70
with the Composition metric rating alone! By adding in Hydrology and Soils into the IEI rating, the
correlation between Ml Condition and EIA Condition drops to a moderate correlation (r = 0.44). Ml
Condition rating and mean CC had no significant correlation (r = 0.12), whereas EIA Vegetation Rating
and mean CC had a moderate correlation (r = 0.49).
For the HSI rating, the only Ml factor showing any correlation with the stressor index was Ml Landscape
Context (r = 0.49). The overall Ml EORANK had low correlation (r = 0.28). By comparison, IEI Rating had
a moderate correlation (r = 0.47). So, our EIA method assesses a broader range of ecological attributes,
and these show a response to stressor impacts.
APPLICATION OF IEI SCORES TO WETLAND SITES
We used our final form of the IEI, along with ratings for each of the Major Attributes to summarize the
reference gradient for wetlands in Ml and IN (Table C13). We summarize the ecological characteristics,
the ecological integrity ratings, and stressors across the major wetland types in terms of Excellent (A),
Good (B), Fair (C), and Poor (D) wetlands.
90
-------
TABLE C13. Summary of the number of sites by wetland macrogroups and a) Index of Ecological
Integrity rating, and b) Condition. Highlighted cells are those with less than 5 replicates.
Macrogroups
Northern & Central Floodplain Forest
Northern & Central Swamp Forest
North American Boreal Bog & Fen
Great Plains Freshwater Wet Meadow, Wet Prairie & Marsh
Appalachian, Interior Plateau & Prairie Fen
Eastern North American Wet Shrub, Meadow & Marsh
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow
Total
Index of Ecological
Integrity
A
8
5
12
6
6
6
17
60
B
19
21
17
12
32
31
16
148
C
5
5
0
8
2
7
3
30
CD
6
2
0
7
2
9
2
28
D
1
1
1
5
2
1
0
11
Total
39
34
30
38
44
54
38
111
Macrogroups
Northern & Central Floodplain Forest
Northern & Central Swamp Forest
Appalachian, Interior Plateau & Prairie Fen
North American Boreal Bog & Fen
Atlantic & Gulf Coastal Plain Pondshore & Wet Meadow
Eastern North American Wet Shrub, Meadow & Marsh
Great Plains Freshwater Wet Meadow, Wet Prairie & Marsh
Total
Condition
A
17
17
20
24
26
12
11
127
B
21
14
21
5
9
36
20
126
C
1
2
3
1
3
7
6
23
D
1
1
Total
39
33
44
30
38
55
38
111
91
-------
C.4 DISCUSSION
THE ECOLOGICAL INTEGRITY MODEL
Our overall model for wetlands was developed from a large body of wetland condition assessment
methods. Thus the choice of metrics and the variants needed for various wetland types (bogs, marshes,
floodplain forests) was based on previous testing. Being something of a synthesis of other models, we
were concerned that we might have included too many metrics, or that some metrics were not as
relevant to assessing integrity as they might be to other aspects of ecosystem structure and function. In
addition, we included metrics, such as Vegetation Composition, that are less quantitative, and require
greater ecological experience than other metrics, such as Percent cover of Natives. For that reason, we
felt it was important to check the model for redundancy and discriminatory power.
By-and-large, the individual metrics, as well as aggregate major ecological factors,primary rank factors,
and the overall index of ecological integrity showed little redundancy and good discriminatory power.
Still, of the 18 major metrics we included, we felt we could do with as few as 14 (one to four metrics per
six major attributes). Reducing the number of metrics will increase the efficiency of the overall process.
In evaluating the contribution of various ecological attributes to understanding ecological integrity, we
found that Vegetation was not a simple proxy for Hydrology or Soils, or does not respond strongly to
abiotic stressors (Table Cll). Our tests used a Human Stressor Index based on Stressors to the abiotic
(Hydrology and Soils) and buffer attributes, but the vegetation metrics responded somewhat weakly to
those stressors (r = 0.45). Changes in the vegetation may reflect other stressors (e.g. logging history,
deer browse) not assessed by the on-site abiotic stressors, or reflect longer term response to stressors
that are not currently evident on the site. The relative weight given to the Vegetation attribute may
vary depending on the application. From a biodiversity perspective, some may feel that, even if
hydrology and soils are not degraded, but vegetation is strongly degraded, then the IEI should be rated
as poor. From a functional perspective more weight could be given to hydrology or soils. Here we chose
a balanced set of weightings.
Our model, more than some, includes metrics that require experienced wetland ecologists to properly
rate them, much as wetland delineation requires experienced evaluators. We believe the use of such
metrics to be the strength of the method, allowing us to retain at least a four point rating of integrity (A,
B, C, D), as opposed to A/B, C and D.
Our field assessments rely on observable features at the site. But past land use history may play an
important role in shaping current condition. This includes activities that have occurred (or not) on a
particular site anywhere from a few years ago to 200 years ago. Their influence may be unknown or very
uncertain to, and yet they may affect the level of ecological integrity. For example, a site may have
altered hydrology, but when combined with decades of very benign land use, they maintain a good
condition. There may also be lag effects from land uses. Impacts from crop production (edge effects
from pesticides, nitrogen enrichment and runoff) may show up right away in condition, whereas fire
suppression might not show up strongly until a tipping point is reached, maybe after decades or even
100 years.
92
-------
OVERALL LEVEL 2 INDEX AND LEVEL 3 METRICS
By constructing a model that consistently assesses the major attributes of a wetland, we are able to
identify those aspects of the wetland that are in better or worse shape. Thus, we are able to show here
that few wetlands have soil disturbances, but many have altered hydrology and vegetation. Integrating
these various attributes into an overall rank provides a concise and readily interpretable rating of
wetland integrity for wetland managers, conservationists, and the public.
The IEI is based on the perspective of a minimally disturbed reference condition, based on historical
integrity of the site, lack of negative human impacts, and a surrounding landscape dominated by natural
ecosystems and processes. But a high score for integrity may not necessarily translate into high scores
for ecosystem services. That is a separate evaluation. We caution that wetlands may be in excellent
condition but not may not be considered high scoring for any given ecosystem service. Thus, ecosystem
services should be evaluated based on both the inherent capacity of natural ecosystems as well as
potential capacity based on modifications to those systems. For example, floodplain forests with high
ecological integrity have a range in capacity for providing flood control services; these forests could also
be modified to increase those services, but depending on the modification, this may or may not
maintain their level of integrity.
Although our weighting of metrics is general, it has the advantage that it can be applied to any wetland,
at least in the temperate and boreal regions. There are advantages to fine-tuning the model, weighting
some metrics or attributes higher than others, depending on the wetland. For example, Jacobs et al.
(2010) found some justification for weighting buffer more strongly than hydrology for depressional
wetlands, and vice-versa for riverine and flats. Nonetheless, applying these rules by wetland type
resulted in only minimal gains in the sensitivity of the index, using intensive quantitative metrics (Jacobs
et al. 2010, Table 5). Thus for rapid assessments, we feel a general model will suffice, with Condition
weighted 60% (Vegetation 24, Hydrology 24, and Soils 12), Size 15%, and Landscape Context 25%. Thus
all major attributes are scored roughly in the 12 - 25% range. This is in keeping with the IEI model, in
that these attributes are included precisely because they reflect major attributes of the wetland, and as
our data show, are each responsive to stressors.
The Level 3 metric based on Coefficient of Conservatism added a little to our understanding of
ecological integrity, particularly in validating our Level 2 Vegetation metric (r = 0.61). Among more
specific Vegetation metrics, Mean CC showed the highest correlation with Composition (r = 0.50) and
regeneration (r = 0.53) metrics, suggesting some support for the expert evaluation of vegetation.
Further analyses of the Level 3 data are needed. Also, though not the purpose of this study I think the
following is true: Level 3 assessments serve to enhance field skills of both beginner and experienced
level surveyors; by requiring relatively comprehensive plant species identification and cover estimates,
those skills are honed. Conducting the advanced work of such plot studies tend to create stronger
interpretive skills for the less advanced Level 2 surveys.
93
-------
CALIBRATING IEI WITH REMOTE SENSING MODELS.
The Level 1 integrity ranks are often used as a means of prioritizing sites for field visits where Level 2 or
Level 3 assessments will be completed (e.g., see Fennessy et al. 2007), and ranks based on those
assessments would supersede Level 1 ranks. Thus Level 1 assessments can be informative about the
overall range in conditions across a population of wetlands in a landscape or region. They can serve as a
helpful screening method for identifying the most likely conditions on the ground. Level 1 ratings can
also be used as predictors of Level 2 or 3 ratings at individual sites. Tests completed to date, however,
show that Level 1 methods do not accurately predict individual site ratings, particularly on-site
conditions (Mack 2006, Fennessy et al. 2007). Our tests for wetland sites in Michigan and Indiana bear
this out (Fig. 8). However, these tests shows that our methods are more successful in predicting overall
IEI scores, because landscape context and size, as well as on-site condition are part of the IEI, and these
can be effectively assessed using Level 1 metrics. It may also be possible to re-calibrate the metrics used
for Level 1 assessments based on these Level 2 scores.
STUDY DESIGN-ASSESSING WETLANDS VERSUS POINTS.
Our sampling of the reference gradient (minimally disturbed to degraded) was designed to assess
"wetlands," versus a fixed area around the sample point. In the context of ambient monitoring this
decision may be less desirable, at least from some aspects. As Fennessy et al. (2007) note, probabilistic
surveys that have been undertaken have taken an area-based approach rather than assessing a
"wetland." This approach avoids 1) the need for determining an assessment unit boundary (which can
become difficult in large contiguous complexes of wetlands), and 2) measuring the area of the
assessment unit. It also allows points to fall onto disturbed and undisturbed areas of wetlands and be
separately assessed, which avoids having multiple sample points being dropped on the same "wetland"
(since the available digital sample frames will probably not correspond to assessment units defined by
the assessment unit rules of the sample protocol). But there are also several distinct advantages to using
the polygon approach, as noted by Fennessy et al. (2007): the basic "currency" in Clean Water Act
Section 401/404 regulation of wetlands is something called a "wetland" and this is understood to be "a
definable piece of real estate that can be mapped and walked around." These advantages are also
evident in many conservation and resource management contexts and why the Natural Heritage
methodology routinely requires mapping of ecosystem occurrences.
We found that defining an assessment unit boundary was relatively straightforward at the majority of
sites, since most wetlands in the watershed were relatively small in size (< 50 ha). We were able to
obtain digitized assessment unit boundaries and area estimates in advance, but occasionally crews came
to a site wherein natural community type boundaries, natural community quality boundaries, and
jurisdictional wetland boundaries were in different locations, and had to draw the assessment unit
directly onto the aerial photo and digitize this by hand in ArcView. Our field manual will need to
improve the guidance on defining Assessment Areas to deal with these complex situations (Faber-
Langendoen 2011). But given the advantages of assessing wetlands based on polygons, rather than
points, we encourage future assessments to use them.
A STANDARD METHOD FOR ASSESSING WETLAND CONDITION
Building on the work of other rapid assessment methods (ORAM, CRAM), we show here that ecological
integrity can be effectively assessed using a suite of rapid assessment metrics, structured around a
general ecological model. Although some of our metrics require greater expertise than others, all
attributes have at least 2 metrics that can be evaluated in a relatively straightforward manner, allowing
for wide applicability. Many of these metrics are comparable to the draft metrics being tested as part of
the USA-RAM (Table A8), suggesting that a standardized rapid assessment method for wetlands can be
94
-------
achieved. This method will have great value for the Natural Heritage Network as well, as an improved
method for assessing wetland condition within and among states and provinces.
REFERENCES
Andreasen, J.K., R.V. O'Neill, R. Noss, and N. C. Slosser. 2001. Considerations for the development of a
terrestrial index of ecological integrity. Ecological Indicators 1: 21-35.
Bailey, R.G. 1996. Ecosystem Geography. Springer-Verlag, New York.
Barbour, M.T., J. Gerritsen, G.E. Griffith, R. Frydenborg, E. McCarron, J.S. White, and M.L Bastian. 1996.
A framework for biological criteria for Florida streams using benthic macroinvertebrates. Journal of
North American Benthological Society 15(2): 185-211.
Beyer, H. L 2004. Hawth's Analysis Tools for ArcGIS. Available at http://www.spatialecology.com/htools
Blocksum, K.A. 2003. A performance comparison of metric scoring methods for a multi-metric index for
mid-Atlantic Highlands streams. Environmental Management 31: 670-682.
Blocksom, K. A., J. P. Kurtenbach, D. J. Klemm, F. A. Fulk, and S. M. Cormier. 2002. Development and
Evaluation of the Lake Macroinvertebrate Integrity Index (LMII) for New Jersey Lakes and Reservoirs.
Environmental Monitoring and Assessment 77:311-333.
Brewer, J.S. and T. Menzel. 2009. A method for evaluating outcomes of restoration when no reference
sites exist. Restoration Ecology 17:4-11.
Brinson, M. M. 1993. A Hydrogeomorphic Approach to Wetland Functional Assessment. Technical
Report WRP-DE-4. Waterways Experiment Station, U.S. Army Corps of Engineers, Vicksburg, MS.
Brooks, R.P., D.H. Wardrop, and J.A. Bishop. 2004. Assessing wetland condition on a watershed basis in
the Mid-Atlantic region using synoptic land-cover maps. Environmental Monitoring and Assessment
94:9-22.
Brown,N., L. Master, D. Faber-Langendoen, P. Comer, K. Maybury, M. Robles, J. Nichols and T.B. Wigley.
2004. Managing elements of biodiversity in sustainable forestry programs: Status and utility of
NatureServe's information resources to forest managers. Technical Bulletin No. 885. Research
Triangle Park, N.C.: National Council for Air and Stream Improvement, Inc.
Collins, J.N., E.D. Stein, M. Sutula, R. Clark, A.E. Fetscher, L. Grenier, C. Grosso, and A. Wiskind. 2006.
California Rapid Assessment Method (CRAM) for Wetlands and Riparian Areas. Version 4.2.3. 136
pp.
Comer, P., D. Faber-Langendoen, R. Evans, S. Gawler, C. Josse, G. Kittel, S. Menard, M. Pyne, M. Reid, K.
Schulz, K. Snow, and J. Teague. 2003. Ecological Systems of the United States: A Working
Classification of U.S. Terrestrial Systems. NatureServe, Arlington, VA.
95
-------
Comer, P., and K. Schulz. 2007. Standardized Ecological Classification for Meso-Scale Mapping in
Southwest United States. Range/and Ecology and Management 60 (3) 324-335.
Comer, P.J. and J. Hak. 2009. NatureServe Landscape Condition Model. Internal documentation for
NatureServe Vista decision support software engineering, prepared by NatureServe, Boulder CO.
[draft]
Cowardin, L M., V. Carter, F. C. Golet, and E. T. LaRoe. 1979. Classification of the wetlands and
deepwater habitats of the United States. U.S. Fish and Wildlife Service, Washington, D.C., USA
Danz, N.P., G.J. Neimi, R.R.Regal. et al. 2007. Integrated measures of anthropogenic stress in the U.S.
Great Lakes Basin. Environmental Management 39:631-647.
DeKeyser, E.S., D.R. Kirby, and M.J. Ell. 2003. An index of plant community integrity: development of the
methodology for assessing prairie wetland plant communities. Ecological Metrics 3: 119-133.
De Leo, G. A., and S. Levin. 1997. The multifaceted aspects of ecosystem integrity. Conservation Ecology
[online] l:3:Available from http://www.consecol.org/voll/issl/art3.
Egan, D. and E.A. Howell. 2005. Introduction. Pp 1-23 In D. Egan and E.A. Howell. The Historical Ecology
Handbook, Island Press, WA.
Faber-Langendoen, D., G. Kudray, C. Nordman, L. Sneddon, L. Vance, E. Byers, J. Rocchio, S. Gawler, G.
Kittel, S. Menard, P. Comer, E. Muldavin, M. Schafale, T. Foti, C. Josse, J. Christy. 2008. Ecological
Performance Standards for Wetland Mitigation based on Ecological Integrity Assessments.
NatureServe, Arlington, VA. + Appendices.
Faber-Langendoen, D, J. Rocchio, P. Comer, G. Kudray, L. Vance, E. Byers, M. Schafale, C. Nordman, E.
Muldavin, G. Kittel, L. Sneddon, M. Pyne and S. Menard. 2008. Overview of Natural Heritage
Methodology for Ecological Element Occurrence Ranking based on Ecological Integrity Assessment
Methods [Draft for Network Review]. NatureServe, Arlington, VA.
Faber-Langendoen, D., D.LTart, and R.H.Crawford. 2009a. Contours of the revised U.S. National
Vegetation Classification standard. Bulletin of the Ecological Society of America 90:87-93.
Faber-Langendoen, D, R. Lyons, and P. Comer. 2009b. Developing options for establishing reference
conditions for wetlands across the lower 48 states. A report to the U.S. Environmental Protection
Agency. NatureServe, Arlington, VA.
Faber-Langendoen, D., C. Hedge, M. Kost, S. Thomas, L. Smart, R. Smyth, J. Drake, and S.
Menard. 2011. Assessment of wetland ecosystem condition across landscape regions: A
multi-metric approach. NatureServe, Arlington VA.+ Appendices.
Faber-Langendoen, D., J. Rocchio, S. Thomas, M. Kost, C. Hedge, B. Nichols, K. Walz, G. Kittel, S. Menard,
J. Drake, and E. Muldavin. 2012b. Assessment of wetland ecosystem condition across landscape
regions: A multi-metric approach. Part B. Ecological Integrity Assessment protocols for rapid field
methods (12). EPA/600/R-12/021b. U.S. Environmental Protection Agency Office of Research and
Development, Washington, DC.
96
-------
Fancy, S.G., J.E. Gross and S. L Carter. 2009. Monitoring the condition of natural resources in US
national parks. Environ. Monit. Assess. 151:161-174.
Federal Geographic Data Committee. 2008. Vegetation Classification Standard, version 2 FGDC-STD-005,
v2. Washington, DC.
Fennessy, M.S., A.D. Jacobs, and M.E. Kentula. 2007. An evaluation of rapid methods for assessing the
ecological condition of wetlands. Wetland 27:543-560.
Fennessy, M. S., J. J. Mack, E. Deimeke, M. T. Sullivan, J. Bishop, M. Cohen, M. Micacchion and M. Knapp.
2007. Assessment of wetlands in the Cuyahoga River watershed of northeast Ohio. Ohio EPA
Technical Report WET/2007-4. Ohio Environmental Protection Agency, Division of Surface Water,
Wetland Ecology Group, Columbus, Ohio.
Forman, R.T.T. 1995. Land Mosaics: the ecology of landscapes and regions. Cambridge University Press.
Harwell, M.A., V. Myers, T. Young, A. Bartuska, N. Gassman, J. H.Gentile, C. C. Harwell, S. Appelbaum, J.
Barko, B. Causey, C. Johnson, A. McLean, R. Smola, P. Templet, and S. Tosini. 1999. A framework for
an ecosystem integrity report card. BioScience 49: 543-556.
Heinz Center. 2002. The State of the Nation's Ecosystems: Measuring the Lands, Waters and Living
Resources of the United States. The Heinz Center for Science, Economics and the Environment.
Cambridge University Press, New York. 270 pp.
Herman, K.D., LA. Masters, M.R. Penskar, A.A. Reznicek, G.S. Wilhelm, W.W. Brodowicz, and K.P.
Gardiner. 2001. Floristic quality assessment with wetland categories and computer application
programs for the state of Michigan. Michigan Dept. of Natural Resources, Wildlife Division, Natural
Heritage Program, Lansing, Ml.
Herrick J.E., J.W. Van Zee, K.M. Havstad, L. Burkett and W. G. Whitford. 2005. Monitoring manual for
grassland, shrubland and savanna ecosystems. Vol 1: quick start. Tucson, AZ: University of Arizona
Press.
Higgs, E. 2003. Nature by Design; People, Natural Process, and Ecological Restoration. The MIT Press,
Cambridge, Massachussets. 341 p.
Higgs, E.S., and R.J. Hobbs. 2010. Wild design: Principles to guide interventions in protected areas. Pp
234-251 In D.N. Cole and L. Yung. Beyond naturalness: Rethinking park and wilderness stewardship
in an era of rapid change. Island Press, Washington, DC.
Holling C.S. 1973. Resilience and stability of ecological systems. Annual Review of Ecology and
Systematics 4: 1-23.
Homoya, M.A., D.B. Abrell, J.R. Aldrich & T.W. Post. 1985. The natural regions of Indiana. Proceedings of
the Indiana Academy of Science 94:245-268
Hruby, T. 2001. Environmental auditing: testing the basic assumption of the hydrogeomorphic approach
to assessing wetland functions. Environmental Management Vol. 27, No. 5, pp. 749-761.
Huggins, D.G. and A. Dzialowski. 2005. Identification and Quantification of Reference Conditions
Associated with Lotic Ecosystems of the Central Plains and Surrounding Regions: A summary of
97
-------
approaches and factors and regional approach. Report 134 of the Kansas Biological Survey.
University of Kansas, Lawrence, KS.
Jacobs, A.D. 2003. Delaware Rapid Assessment Procedure, Version 1.2. Delaware Department of natural
Resources and Environmental Control. Dover, DE. USA.
Jacobs, A., E. Mclaughlin, and K. Havens. 2010. Mid-Atlantic Tidal Wetland Rapid Assessment Method
Version 3.0. June 2010. Delaware Department of Natural Resources and Environmental Control.
Dover, DE.
Jacquart, E., M. Homoya, L Casebere. 2002. Natural communities of Indiana. July 1, 2002 Working
Draft. Department of Natural Resources, Division of Nature Preserves. Indianaopolis, IN.
Jennings, M.D., D. Faber-Langendoen, O.L Loucks, R.K. Peet, and D. Roberts. 2009. Standards for
associations and alliances of the U.S. National Vegetation Classification. Ecological Monographs 79:
173-199.
Kapos, V, I Lysenko, and R Lesslie. 2002. Assessing forest integrity and naturalness in relation to
biodiversity. Forest Resources Assessment Programme, Working Paper 54, Rome 2002. Forestry
Department Food and Agriculture Organization of the United Nations. FRA 2000.
Karr, J.R. and E.W. Chu. 1999. Restoring life in running waters: better biological monitoring. Washington
(DC). Island Press, 206 pp.
Klimas, Charles V., Elizabeth O. Murray, Henry Langston, Theo Witsell, Thomas Foti, Rob Holbrook. 2006.
A Regional Guidebook for Conducting Functional Assessments of Wetland and Riparian Forests in
the Ouachita Mountains and Crowley's Ridge Regions of Arkansas. ERDC/EL TR-06-14 U.S. Army
Engineer Research and Development Center. Vicksburg, MS
Kost, M.A., D.A. Albert, J.G. Cohen, B.S. Slaughter, R.K, Schillo, C.R. Weber, and K.A. Chapman. 2007.
Natural Communities of Michigan: Classification and Description. Michigan Natural Features
Inventory, Report Number 2007-21, Lansing, Ml 314 pp..
Landres, PB, P Morgan and FJ Swanson. 1999. Overview of the use of natural variability concepts in
managing ecological systems. Ecological Applications 9(4): 1179-1188.
Lapp, H., Morris, R. A., Catapano, T., Hobern, D. and Morrison, N. (2011), Organizing our knowledge of
biodiversity. Bulletin of the American Society for Information Science and Technology, 37: 38-42.
doi:10.1002/bult.2011.1720370411
Lindenmayer, D.B., and J.F. Franklin. 2002. Conserving forest biodiversity: A comprehensive multiscaled
approach. Island Press, Washington, DC. 351 p.
Linkov, I., D. Loney, S. Cormier, F. K. Satterstrom and T. Bridges. 2009. Weight-of-evidence evaluation in
environmental assessment: Review of qualitative and quantitative approaches. Science of the
Total Environment 407:5199-5205.
Mack, J.J., 2001. Ohio rapid assessment method for wetlands v. 5.0, user's Manual and scoring forms.
Ohio EPA Technical Report WET/2001-1. Ohio Environmental Protection Agency, Division of Surface
Water, Wetland Ecology Group, Columbus, Ohio.
98
-------
Mack, J. J. 2004. Integrated wetland assessment program, part 4: Vegetation Index of Biotic Integrity
(VIBI) and Tiered Aquatic Life Uses (TALUs) for Ohio wetlands. Ohio EPA Technical Report WET/2004-
4, Ohio Environmental Protection Agency, Wetland Ecology Group, Division of Surface Water,
Columbus, Ohio.
Mack, J.J. 2006. Landscape as a predictor of wetland condition: an evaluation of the Landscape
Development Index (LDI) with a large reference wetland dataset from Ohio. Environmental
Monitoring and Assessment 120: 221-241.
Mack, J.J. 2007a. Developing a wetland IBI with statewide application after multiple testing iterations.
Ecological Indicators 7: 864-881.
Mack, J.J. 2007b. Integrated wetland assessment program, part 9: Field manual for the Vegetation Index
of Biotic Integrity for wetlands. Ohio EPA Technical Report WET/2007-6, Ohio Environmental
Protection Agency, Wetland Ecology Group, Division of Surface Water, Columbus, Ohio.
Mack, J. J. and M.E. Kentula. 2010. Metric Similarity in Vegetation-Based Wetland Assessment Methods.
EPA/600/R-10/140. U.S. Environmental Protection Agency, Office of Research and Development,
Washington, D.C.
McRae, B.H., B.G. Dickson, T.H. Keitt, and V.B. Shah. 2008. Using circuit theory to model connectivity in
ecology and conservation. Ecology 10: 2712-2724.
McCune, B., and Grace, J.B. 2002. Analysis of ecological communities. MJM Software Design. Gleneden
Beach, Oregon.
McCune, B, and Mefford, M.J. 2006. PC-ORD Multivariate Analysis of Ecological Data, Version 5.0,MjM
Software Design, Gleneden Beach, Oregon.
MDEQ (Michigan Department of Environmental Quality. 2008. Michigan Rapid Assessment Methods for
Wetlands (MiRAM), Version 2.0. MDEQ, Land and Water Management Division, Wetlands, Lakes
and Streams Unit, Lansin, Ml.
Millar Cl, Stephenson NL, Stephens SL (2007) Climate change and forests of the future: managing in the
face of uncertainty. Ecol Appl 17:2145-2151.
Miller, S. J. and D. H. Wardrop. 2006. Adapting the Floristic Quality Assessment Index to reflect
Anthropogenic Disturbance in Central Pennsylvania Wetlands. Ecological Indicators 6:313-26.
Miller, S.J., D.H. Wardrop, W.M. Mahaney, R. P. Brooks. 2006. A plant-based index of biological integrity
(IBI) for headwater wetlands in central Pennsylvania. Ecological Indicators 6:290-312.
Mita, D., E. DeKeyser, D. Kirby, and G. Easson. 2007. Developing a wetland condition prediction model
using landscape structure variability. Wetlands 27:1124-1133.
Mitchell, B. R., G.L. Tierney, K.M. Miller, B. Schweiger, D. Faber-Langendoen, J.B. Grace. 2011 (in press).
Getting the message across: Using ecological integrity to communicate with resource managers. Pp.
xx-xx In Guntenspergen. G. et al. Application of threshold concepts in natural resource decision
making. Springer, New York.
99
-------
NatureServe. 2002. Element Occurrence Data Standard. On-line at
http://www.natureserve.org/prodServices/eodata.isp
NatureServe. 2005. NPS PLOTS Database Version 2.0. NatureServe, Arlington, Virginia.
NatureServe. 2008. International Ecological Classification Standard: Terrestrial Ecological Classifications.
NatureServe Central Databases. Arlington, VA. U.S.A.
NatureServe. 2009. NatureServe National Systems Map. [N.B. Reference incomplete]
NatureServe Vista version 2.0. 2009. www.natureserve.org/vista. NatureServe, Arlington, VA.
NatureServe. 2011. Central Basin and Range. Rapid Ecoregional Assessments, Final Memorandum 1-3-C
(March 4, 2011). Department of the Interior, Bureau of Land Management, Rapid Ecoregional
Assessments, Denver, Colorado. NatureServe, Washington DC.
Noon, B. R. 2003. Conceptual issues in monitoring ecological systems. Pages 27-71 in D. E. Busch and J.
C. Trexler, editors. Monitoring ecosystems: Interdisciplinary approaches for evaluating ecoregional
initiatives. Island Press, Washington, DC., USA 447 pp.
Oakley, K.L., L P. Thomas, and S G. Fancy. 2003. Guidelines for long-term monitoring protocols. Wildlife
Society Bulletin 31(4):1000-1003
Oliver, I., H. Jones, and D.L. Schmoldt. 2007. Expert panel assessment of attributes for natural
variability benchmarks for biodiversity. Austral Ecology 32: 453-475.
Parkes, D., G. Newell and D. Cheal. Assessing the quality of native vegetation:The 'habitat hectares'
approach ECOLOGICAL MANAGEMENT & RESTORATION VOL 4 SUPPLEMENT: 2003 S29-S38.
Parrish, J.D., D. P. Braun, and R.S. Unnasch. 2003. Are we conserving what we say we are? Measuring
ecological integrity within protected areas. BioScience 53: 851-860.
Peet, R. K., T. R. Wentworth, and P. S. White. 1998. The North Carolina Vegetation Survey protocol: a
flexible, multipurpose method for recording vegetation composition and structure. Castanea
63:262-274.
Pellant, M., P. Shaver, D.A. Pyke, and J.E. Herrick. 2005. Interpreting indicators of rangeland health,
version 4. Technical Reference 1734-6. U.S. Department of the Interior, Bureau of Land
Management, National Science and Technology Center, Denver, CO. BLM/WO/ST-
00/001+1734/REV05. 122 pp.
Rocchio, J. 2007. Assessing Ecological Condition of Headwater Wetlands in the Southern Rocky
Mountain Ecoregion Using a Vegetation Index of Biotic Integrity. Colorado Department of Natural
Resources, and U.S. Environmental Protection Agency, Region VIM. Colorado Natural Heritage
Program, Colorado State University, Fort Collins, CO. Online:
http://www.cnhp.colostate.edu/reports.html
Rowe, J.S. and B.V. Barnes. 1994. Geo-ecosystems and bio-ecosystems. Bulletin of the Ecological
Society of America 75 (1): 40-41.
100
-------
Society for Ecological Restoration (SER) International Science & Policy Working Group. 2004. The SER
International Primer on Ecological Restoration, www.ser.org & Tucson: Society for Ecological
Restoration International. 13 p.
Smith, R. D., A. Amman, C. Bartoldus, and M. M. Brinson. 1995. An approach for assessing wetland
functions using hydrogeomorphic classification, reference wetlands, and functional indices.
Technical report TR WRP-DE-10, and operational draft. US Army Engineers Waterways Experiment
Station, Vicksburg MS.
Stein, B.A., LS. Kutner, and J.S. Adams (eds.). 2000. Precious Heritage: The status of biodiversity in the
United States. The Nature Conservancy and Association for Biodiversity Information [now
NatureServe]. Oxford University Press.
Stein, E.D., A. E. Fetscher, R. P. Clark, A. Wiskind, J. L Grenier, M. Sutula, J. N. Collins, and C. Grosso.
2009. Validation of a wetland rapid assessment method: use of EPA's level 1-2-3 framework for
method testing and refinement wetlands. Wetlands 29: 648-665
Stevens, D.L and S.F. Jensen. 2007. Sample design, execution, and analysis for wetland assessment.
Wetlands 27: 515-523.
Stoddard, J.L, D.P. Larsen, C.P. Hawkins, R.K. Johnson, and R.H. Norris. 2006. Setting expectation for
the ecological condition of streams: the concept of reference condition. Ecological Applications
16(4): 1267-1276.
Sutula, M.A., E.D. Stein, J.N. Collins, A.E. Fetscher, and R. Clark. 2006. A practical guide for development
of a wetland assessment method: the California experience. J. Amer. Water Resources Association
42:157-175.
Swetnam, T.W., C.D. Allen, and J.L. Betancourt. 1999. Applied historical ecology: using the past to
manage for the future. Ecological Applications 9:1189-1206.
Tierney, G.L, D. Faber-Langendoen, B.R. Mitchell. G. Shriver, J. Gibbs. 2009. Monitoring and evaluating
the ecological integrity of forest ecosystems. Frontiers in Ecology and the Environment 7: 308-316.
Tierney, G. L, B. R. Mitchell, K. M. Miller, J. Comiskey, A. Kozlowski, and D. Faber-Langendoen. 2010.
Northeast Temperate Network long-term forest monitoring protocol: 2010 revision., National Park
Service, Fort Collins, Colorado.
Tiner, R.W. 2004. Remotely-sensed indicators for monitoring the general condition of "natural habitat"
in watersheds: an application for Delaware's Nanticoke River watershed Ecological Indicators 4
(2004)227-243.
Unnasch, R.S., D. P. Braun, P. J. Comer, G. E. Eckert. 2009. The Ecological Integrity Assessment
Framework: A Framework for Assessing the Ecological Integrity of Biological and Ecological
Resources of the National Park System. Report to the National Park Service.
U.S. Department of Agriculture, Natural Resources Conservation Service (USDA NRCS). 1997. National
range and pasture handbook. Washington, DC., U.S. Department of Agriculture.
U.S. EPA. 2002. Methods for evaluating wetland condition: wetlands classification. Office of Water, U.S.
Environmental Protection Agency, Washington, DC. EPA-822-R-02-017.
101
-------
U.S. EPA. 2006. Application of Elements of a State Water Monitoring and Assessment Program For
Wetlands, Wetlands Division, Office of Wetlands, Oceans and Watersheds, U.S. Environmental
Protection Agency, Washington, DC, April 2006.
USEPA. 2007. Level III Ecoregions of the conterminous United States.
www.epa.gov/wed/pages/ecoregions/level iii.htm
USEPA (U.S. Environmental Protection Agency). 2011. National Wetland Condition Assessment: Field
Operations Manual. EPA-843-R-10-001. U.S. Environmental Protection Agency, Washington, DC.
U.S. Fish and Wildlife Service (USFWS). 2010. NWIPIus: Geospatial Database for Watershed-level
Functional Assessment. April 2010 Factsheet.
http://www.fws.gov/northeast/wetlands/factsheets/NWIPIus FactSheet.pdf
Voss, E. G. 1972. Michigan Flora. Part I. Gymnosperms and Monocots. Cranbrook Institute of Science
Bulletin 55 and the University of Michigan Herbarium, xv + 488 pp.
Voss, E. G. 1985. Michigan Flora. Part II. Dicots (Saururaceae-Cornaceae). Cranbrook Institute of Science
Bulletin 59 and the University of Michigan Herbarium, xix + 724 pp.
Voss, E. G. 1996. Michigan Flora. Part III. Dicots (Pyrolaceae-Compositae). Cranbrook Institute of Science
Bulletin 61 and the University of Michigan Herbarium, xix + 622 pp.
Williams, J.W., ST. Jackson, and J. Kutzbach. 2007. Projected distributions of novel and disappearing
climates by 2100 AD. PNAS 104: 5738-5742
Yorks, T. E., and S. Dabydeen. 1998. Modification of the Whittaker sampling technique to assess plant
diversity in forested natural areas. Natural Areas Journal 18:185-189.
Young, T.F. and S. Sanzone (editors). 2002. A framework for assessing and reporting on ecological
condition. Prepared by the Ecological Reporting Panel, Ecological Processes and Effects Committee.
EPA Science Advisory Board. Washington, DC. 142 p.
102
-------
APPENDICES
Appen dix 1. SAMPLING DESIGN AND ASSESSMENT AREA
1. Sampling Design
A sampling design is often used to implement both a rapid and an intensive wetland monitoring
program; however, creating such a design is beyond the scope of this report. But the sampling
design determines how the field crews navigate to a site. See Section B of the report for our
sampling design, which focused on developing a reference gradient of sites, including reference
standard (minimally disturbed) sites.
2. Determine the extent of a wetland type at the site, and classify the wetland type
Wetlands will be classified using a variety of classifications, including state Natural Heritage types,
the U.S. National Vegetation Classification (FGDC 2008), and NatureServe's Ecological Systems
(Comer et al. 2003). Knowing the Formation and System will determine which metrics are used
and the rating scheme for the metrics, in so far as these vary by formation. For example, assessing
the Hydroperiod of a freshwater marsh requires a different form of the Hydroperiod metric than for
a bog or forested wetland. Preliminary mapping can be done in the office.
3. Establish the Assessment Area
Field methods for applying ecological integrity assessments vary, depending on the purpose of the
assessment. But several general comments can be provided, in the context of typical ground based
rapid assessments. First, the level of inference must be established. Most commonly, for ecological
surveys, this is an occurrence of a wetland, at the scale of a site. We refer to this as the Assessment
Area (AA). Accordingly we may define the AA as "the entire area, sub-area, or point of an
occurrence of an ecosystem type."
If the occurrence or polygon at a site is the focus, then there are multiple possible strategies for
sampling the occurrence:
1) conduct an assessment survey of the entire AA of the occurrence
2) conduct an assessment survey of a typical sub-area(s) of the occurrence, or
3) collect data at a series of plots, placed in representative or un-biased locations, throughout
the entire area or sub-area occurrence.
But it may also be that the focus is simply on the point selected by, e.g., a sampling design, and there
is no intent to make inferences about the entire extent of a wetland occurrence at a site.
4) Conduct an assessment of the point specified by the survey design, with the intent of
inferring condition across a watershed, managed area, or jurisdiction.
Here, our primary focus is that of working at the level of an occurrence; that is, an entire local
wetland polygon of a type with relatively uniform conditions. The goal is to assess the integrity of
103
-------
this occurrence, irrespective of the property or management regime it may be found on, and
however large it is.
4. Conduct Field Survey
The field methods used for ecological integrity assessments required expertise that is akin to that
needed for wetland delineation; that is, field crews should have some knowledge of hydrology, soils,
and vegetation, sufficient to assess hydrologic dynamics, perhaps examine a soil core for mottling,
and be able to identify all prominent exotic species in a region. See Faber-Langendoen et al.
(2012b, Appendix 2 - Field Methods) for additional guidance.
104
-------
Appendix 2. LANDSCAPE CONDITION MODEL
Patrick Comer and Jon Hak
Overview
We used NatureServe's Landscape Condition model (Comer and Hak 2009), which uses an
aggregated set of anthropogenic stressors to predict ecosystem condition. It is similar to the
Landscape Development Index used by Mack (2006) and the anthropogenic stress model of Danz et
al. (2007). The model can characterize the entire landscape in terms of the level of stressors
operating at each polygon or pixel. The model could be calibrated in various landscapes, based on
comparing predicted results to ground observations, so that only stressors known to affect
condition/integrity of the ecosystems being studied are used in the model. Here we use it in its
general form, to compare it against a specific set of Landscape Context Metrics.
The algorithm for the NatureServe model integrates various land use CIS layers (roads, land cover,
water diversions, groundwater wells, dams, mines, etc.) at a 30-90 m or 1 km pixel scale. These
layers are the basis for various metrics, which are based on stressors. The metrics are weighted
according to their perceived impact on ecological integrity, into a distance-based, decay function to
determine what effect these stressors have on landscape integrity. The result is that each grid-cell
(30 m or more) is assigned a stressor "score". The product is a landscape or watershed map
depicting areas according to their potential "integrity." We can segment the index into four rank
classes, from Excellent (slightly impacted) to Poor (highly impacted) (Fig. A2.1). A series of
wetland sites or EOs can then be overlaid on the landscape model, and the level of anthropogenic
stress on those sites can be modeled. For each wetland polygon, a mean stressor score will be
calculated from the model using the zonal statistics tool in ArcView 9.3. When applied to individual
natural occurrences of wetland types, the model scores can be interpreted as predicting the edge
effect based on the strength of stressors around a given occurrence.
105
-------
Figure A2.1 Landscape Condition Model integrating stressors within a watershed (Comer and Hak 2009).
Figure adapted from Rocchio (2007).
A. Introduction
In developing Ecological integrity assessments, we can develop an approach that assesses two
kinds of attributes - the first are attributes of the ecosystem itself, the second are stressors acting
on those attributes. For the first approach, we want to identify a core set of metrics that best
distinguish a highly impacted or degraded state from a relatively unimpaired or intact state, based
on assessing the key ecological attributes (or more general ecological factors). Metrics may be
based either on characteristics that typify a particular ecosystem or attributes that change
predictably in response to anthropogenic stress.
Second, we need to identify attributes that reflect the level of stressors that may be impacting the
condition of the system, and which may be driving changes in these ecological attributes. Where we
can develop a correlation between these two sets of attributes, we can develop a predictive model
of how stressors impact the ecological integrity of the system. In this way, indicators from the first
approach will indicate the magnitude of key stressors acting upon the system and increase our
understanding of relationships between stressors and effects (Tierney et al. 2009).
There are growing sets of information on various kinds of stressors that impact ecosystems. Danz
et al. (2007) noted that "Integrated, quantitative expressions of anthropogenic stress over large
geographic regions can be valuable tools in environmental research and management." When they
take the form of a map, or spatial model, these tools initially characterize ecological conditions on
106
-------
the ground; from highly disturbed to apparently unaltered conditions. They can be particularly
helpful for screening candidate reference sites; i.e., a set of sites where anthropogenic stressors
range from low to high. Ecological condition of reference sites are further characterized to
determine how ecological attributes are responding to apparent stressors. This knowledge may
then apply in other similar sites.
Anthropogenic stressors come in many forms, from regional patterns of acid deposition or climate-
induced ecosystem change, to local-scale patterns in agricultural drainage ditches and tiles, point-
source pollution, land-conversion, and transportation corridors, among others. To be effective, a
landscape condition model needs to incorporate multiple stressors, their varying individual
intensities, the combined and cumulative effect of those stressors, and if possible, some measure of
distance away from each stressor where negative effects remain likely. Since our knowledge of
natural ecosystems is varied and often limited, a primary challenge is to identify those stressors
that likely have the most degrading effects on ecosystems or species of interest. A second challenge
is to acquire mapped information that realistically portrays those stressors. In addition, there are
tradeoffs in costs, complexity, the often varying spatial resolutions in available maps, and the
variable ways stressors operate across diverse land and waterscapes. Typically, expert knowledge
forms the basis of stressor selection, and relative weighting. Once models are developed, they may
be calibrated with field measurements. Developing empirical relationships between stress
variables and ecological response variables is a key to providing insights into how human activities
impact ecological condition (Danz et al. 2007).
B. Landscape Condition Model
There are two primary uses for NatureServe's landscape condition model: 1) to map the predicted
ecological conditions one would encounter in the field, based on apparent stressors present across the
landscape of interest, and 2) facilitate repeated predictions of ecological condition within the same
landscape over time, or given alternative land use proposals. Maps predicting relative ecological
condition can provide a screening tool for gauging anthropogenic stress in locations including any
mapped point or polygon. Repeated predictions of ecological condition assist with evaluating likely
effects of changes in overlapping land uses on the condition of the landscape for an element or
group of elements. This can provide a powerful tool understanding cumulative effects of land use
change over time and/or for modeling environmental restoration options. The landscape condition
model is integrated into NatureServe's Vista software (NatureServe 2009).
C. METHODS
Here we focus on the methods for developing a landscape condition model. This model is needed as
a predictive tool to screen candidate reference sites. The model needs to provide a set of sites that
contain the range of ecological condition (perhaps categorized by High, Moderate, or Low
Condition). At the outset, we use a general set of stressors, presuming that they are relevant to
what's affecting condition on the ground. Ultimately, we would like to calibrate the model with a
robust sampling of field observations so that all model inputs and settings most efficiently reflect
field conditions.
107
-------
C.I. Study Area
The study area is the conterminous U.S, but NatureServe is working with partners to develop
regional models and groups of elements models (e.g., wetlands).
C.2. Selected Stressors
For this national model, we selected a limited set of stress-inducing land use classes for which we
have nationally consistent coverage (Table 1). Our aim here is to characterize the primary local
scale stressors. We have not attempted to factor in regional stressors, such as air pollutants or
climate change. Stressors are organized into thematic groupings of Transportation, Urban and
Industrial Development, and Managed & Modified Land Cover.
TABLE 1. Stressors selected and mapped for modeling landscape condition nationally.
Theme
Transportation
Primary Highways with limited access
Primary Highways without limited
access
Secondary and connecting roads
Local, neighborhood and connecting
roads
Urban and Industrial Development
High Density Developed
Medium Density Development
Low Density Development
Managed & Modified Land Cover
Cultivated Agriculture
Pasture & Hay
Managed Tree Plantations
Introduced Upland Herbaceous
Source
ESRI® Data & Maps: StreetMap™ Series
issue: 2006 United States
ESRI® Data & Maps: StreetMap™ Series
issue: 2006 United States
ESRI® Data & Maps: StreetMap™ Series
issue: 2006 United States
ESRI® Data & Maps: StreetMap™ Series
issue: 2006 United States
National Land Cover Data/ LANDFIRE
Existing Vegetation
200 1-2003 United States
National Land Cover Data/ LANDFIRE
Existing Vegetation
200 1-2003 United States
National Land Cover Data/ LANDFIRE
Existing Vegetation
2001-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
200 1-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
2001-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
200 1-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
Spatial
Resolution
1:100,000
1:100,000
1:100,000
1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
108
-------
Introduced Wetland Vegetation
Introduced Tree & Shrub
Recently Logged
Native Vegetation with Introduced
Species
Ruderal Forest & Upland
200 1-2003 United States |
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
2001-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
200 1-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
2001-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
2001-2003 United States
National Land Cover Data/ LANDFIRE Existing
Vegetation/Gap Analysis Program
200 1-2003 United States
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
30m pixel/ 1:100,000
Transportation features, derived from ESRI streetmap data circa 2006, depict roads of four distinct
sizes and expected traffic volume. These data provide a practical measure of human population
centers and primary transportation networks that link those centers. Ecological stress induced by
built infrastructure (through habitat loss, fragmentation, altered ecological processes, etc.) are well
known.
As a compliment to Transportation infrastructure, Urban and Industrial Development includes
industrial (e.g., mines) and built infrastructure across a range of densities, from high density urban
and industrial zones, to suburban residential development, to exurban residential and urban open
spaces (golf courses, for outdoor recreation. These data were derived from national land cover data
through combined efforts of US Geological Survey (National Land Cover and Gap Analysis
Programs) and the inter-agency LANDFIRE effort
The third category, Managed and Modified Land Cover, includes the gradient of land cover types
that reflect land use stressors at varying intensities. Again, national data from USGS and LANDFIRE
provide a consistent depiction of these varying land cover classes, from intensive (cultivated
and/or irrigated) agriculture, pasture & hay fields, vineyards and timber tree plantations, various
forms of introduced non-native vegetation in upland and wetland environments, and finally, areas
where native vegetation predominates, but modifications have clearly taken place. These
modifications include recently logged areas, or areas that have seen historic conversion, but have
recovered some combination of mainly native vegetation (old fields, 'off-site' hardwoods and
conifers in many southeastern forest, etc.).
C.3. Model Parameters
Relative Site Intensity
Each land cover category was given a relative site intensity score, between 0.0 and 1.0, to
represent our assumptions of stress induced by each land cover type on terrestrial ecological
109
-------
systems and habitat for native species. As depicted in Table 2, a relative site intensity score that is
close to 0.0 indicates our assumption that this land cover induces very high levels of stress on local
ecosystems. Scores closer to 1.0 are assumed to induce some level of stress, but that stress is
much more limited. Generally, each land cover category is listed within these generalized
categories of assumed stress, but their individual numerical scores were used in modeling.
Typically, only one land cover occurs at each pixel, but where more than one can occur, the lowest
score is applied (i.e., the highest-impact use determines the pixel value). Therefore, in instances
where e.g., a roads layer is distinct from the land cover layer, the roads layer could indicate a 0.05
score, and the land cover layer would also provide a 0.05 score for 'high intensity developed.' Only
one value of 0.05 would apply to that pixel.
TABLE 2. Relative Site Intensity scores used for modeling landscape condition nationally.
Theme
Transportation
Primary Highways with limited access
Primary Highways without limited access
Secondary and connecting roads
Local, neighborhood and connecting roads
Urban and Industrial Development
High Density Developed
Medium Density Development
Low Density Development
Managed & Modified Land Cover
Cultivated Agriculture
Pasture & Hay
Managed Tree Plantations
Introduced Upland Herbaceous
Introduced Wetland Vegetation
Introduced Tree & Shrub
Recently Logged
Native Vegetation with Introduced Species
Ruderal Forest & Upland Old Field
Relative Site Intensity
(0.0-1.0)
0.05
0.05
0.2
0.5
0.05
0.5
0.6
0.3
0.9
0.8
0.5
0.3
0.5
0.9
0.9
0.9
Relative Stress at
Point of Impact
Very High
Very High
High
Medium
Very High
Medium
Medium
High
Low
Low
Medium
High
Medium
Low
Low
Low
The site intensity scores attempt to represent the relative degree of ecological stress induced
locally in the immediate area where the land cover occurs. We treat distance effects surrounding
the impacting land cover as a separate component of the model. However, the spatial model will
110
-------
calculate an initial distance effect that varies with the site intensity score of the model. Figure 1
illustrates this initial distance effect resulting solely from the site intensity score of each land cover.
This effect decays to zero within distances ranging from 200-800 meters from the impacting land
cover.
Low
Very High
•WtO.5
•WtO.05
• Wt 0.005
• No Wt
234567
Distance from Site (100m)
FIGURE 1. Default distance effect of site intensity score on initial condition of land cover
type. Here site intensity score is labeled as no Wt ( score of 1.0), Wt 0.5, Wt 0.05, Wt 0.005.
Distance Decay Function
Each land cover category was also given a distance decay function, also scaled between 0.0 and
1.0, to represent our assumptions of decreasing stress-effects of each land cover with distance away
from each impacting feature. The function changes the slope of the initial site intensity curve (Fig.
1) by pushing the terminus of the curve further from the land cover source causing a more gradual
decay to occur. When combined with the site intensity, the decay function may be heavily modified
to represent land cover types such as 4-lane highways where the assumed stress at the site is high
and the distance effect from the feature is long. So, if the site intensity score is low - for high stress
(e.g., 0.3) and the distance decay function is relatively high (e.g., 1.0), the resulting spatial model
would depict the circumstance where the effect of the high stress land cover is expected to decrease
rapidly over short distances. Conversely, if for the same site intensity score (again, 0.3) was given a
low distance decay function (also 0.3) the expected distance effect of that land cover would extend
out over a greater distance.
As depicted in Table 3, a distance decay function that is close to 0.0 indicates our assumption that
this land cover induces very high levels of stress on local ecosystems. Scores closer to 1.0 are
assumed to induce some level of stress, but that stress is much more limited. Generally, each land
cover category is listed within these generalized categories of assumed stress, but their individual
111
-------
numerical scores were used in modeling. Here, we are attempting to represent that relative degree
of ecological stress induced locally in the immediate area where the land cover occurs.
TABLE 3. Distance Decay Functions used for modeling landscape condition nationally.
Theme
Transportation
Primary Highways with limited access
Primary Highways without limited access
Secondary and connecting roads
Local, neighborhood and connecting roads
Urban and Industrial Development
High Density Developed
Medium Density Development
Low Density Development
Managed & Modified Land Cover
Cultivated Agriculture
Pasture & Hay
Managed Tree Plantations
Introduced Upland grass & forb
Introduced Wetland Vegetation
Introduced Tree & Shrub
Recently Logged
Native Vegetation with Introduced
Ruderal Forest & Upland Old Field
Distance Decay Function (0.0-1.0)
0.05
0.05
0.2
0.5
0.05
0.5
0.5
0.5
0.9
0.5
0.5
0.8
0.5
0.5
1.0
1.0
The values applied to the distance effect follow a standardized curve displayed below. For example,
a value of 0.1 represents a distance weight of 1km, and a value of 0.5 is equivalent to 100 meters
(Figure 2).
112
-------
D-Affect
• D-Affect
0 0.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Distance Intensity
Figure 2. Distance value curve used to define condition model.
The default value for both the site intensity and distance intensity are set to zero (0) and any
landcover types with at least one zero value will not be included in the model. For a land cover type
to be included, the user must set intensity values between >0 and 1. If the value is set to 1, the land
cover will be treated as a stressor included in the landscape condition model, but no weight
modifiers will be applied to the land cover type.
The overall intensity at a pixel unit represents the additive combination of all the land cover types
which may overlap at a single pixel. By adjusting the Site Intensity value of a land cover type the
model can be adjusted to cause a land cover type to project a poor landscape condition value across
its overall extent. Because the final landscape condition model surface is a relative index, it is
possible that if the Site Intensity value is not weighted correctly the feature will fail to project poor
landscape condition across its extent.
Because this model works in an additive progression with the final summary normalized against
the maximum value of 1.0, the effect of any one site intensity could be negated by the inclusion of
other land uses with less weight.
The user may need to perform a series of trial runs to insure a land cover is affecting condition
properly in a model. For example, a complex landscape condition model may contain 10 or more
land cover types and in order to insure that a two-lane highway for example, remains a significant
impact, the site intensity value may need to be set at 0.05 or less.
113
-------
The Boost Factor is a power function applied to the final condition model to adjust the distribution
of the model results. The option defaults to a value of 1 and it is recommended that the user only
adjust this value with substantial understanding of how the results will be transformed. Each land
cover type may represent a single land cover or may include multiple sub categories represented by
land cover types preceded by a + symbol.
D. Results
We use an expert-based judgment to compile the layers and create an overall Landscape Condition
Value. In the future, we may use a principal components analysis to integrate the information
from the multiple stressors.
E. Model Evaluation
E.I. Details of Landscape Condition Models in Vista
The starting condition values for an element are used to depict its relative condition across its
distribution which is incorporated into the element ECL. This output is useful for understanding
where the element is in good condition and where it is not, particularly when combined with the
use of a condition threshold. The interpretation of what constitutes a poor landscape condition for
an individual element is strictly up to the user's interpretation of how land covers should affect the
condition of the element on and off site.
114
-------
For setting condition thresholds, observed condition and viability standards for specific elements
can be informative. For instance, a user's goal may be to maintain all element occurrences with EO
Rank values of C or better. As such, the condition value threshold would be set to the Vista
equivalent of 0.5.
Figure 1. Interpretation of landscape condition thresholds using Ecological Integrity (Element
Occurrence Rank) values from Natural Heritage Program databases.
(J
Hi
All Community EOs (CO, GA, IN, LA, ME, Ml, NY NY, WA and WY)
very good
good
fair
poor
AB B BC C
(n=774) (n=1426) (n=51Q) (n=833)
Approximate EO Rank
IndudesallEOsforGA LA, ME, NY, WY, IN&CO,
EOs ONLY from Ml & WA
E.2. Limitations
The concept of landscape condition modeling is highly simplified resulting in relative indices of
condition that take into account a fairly narrow set of considerations. Although experts building and
documenting the model may consider a number of factors in assigning site and distance intensity
weights, the model does not explicitly address issues such as impacts on species mobility,
demographics, habitat connectivity among multiple resources, etc. Other modeling tools exist that
consider some of these issues when knowledge, time, and funding exist to address them.
F. REFERENCES
115
-------
Comer, P.J. and J. Hak. 2009. NatureServe Landscape Condition Model. Internal documentation for
NatureServe Vista decision support software engineering, prepared by NatureServe, Boulder
CO.
Danz, N.P., G.J. Neimi, R.R.Regal. et al. 2007. Integrated measures of anthropogenic stress in the U.S.
Great Lakes Basin. Environmental Management 39:631-647.
Farig, L. and T. Rytwinski. 2009. Effects of Roads on Animal Abundance: an Empirical Review and
Synthesis. Ecology and Society 14(1): 21 (online).
http://www.ecologyandsociety.org/voll4/issl/art21/
NatureServe Vista version 2.0. 2009. www.natureserve.org/vista. NatureServe, Arlington, VA.
116
-------
Appendix 3. LEVEL 2 PROTOCOLS
NatureServe and the Natural Heritage Network have been developing Ecological Integrity Assessments
protocols for rapid field assessments (L2) for a number of years. Here we summarize the various
versions.
Version 1
Version 1 was published in 2008:
Faber-Langendoen,D., G. Kudray, C. Nordman, L Sneddon, L Vance, E. Byers, J. Rocchio, S. Gawler,
G. Kittel, S. Menard, P. Comer, E. Muldavin, M. Schafale, T. Foti, C. Josse, and J. Christy. 2008.
Ecological Performance Standards for Wetland Mitigation based on Ecological Integrity Assessments.
NatureServe, Arlington, VA. + Appendices.
Version 2
Version 2 was published in 2011 as an Appendix C in the first edition of our study in Indiana and
Michigan.
Faber-Langendoen, D., C. Hedge, M. Kost, S. Thomas, L. Smart, R. Smyth, J. Drake, and S. Menard.
2011. Assessment of wetland ecosystem condition across landscape regions: A multi-metric
approach. NatureServe, Arlington VA. + Appendices.
It is the version that guided our field methods and which we tested with field data.
Version 3
Here, in part B of this publication, we are publishing version 3. This is an improved version that
reflects the results of our Michigan and Indiana study, but has been upgraded for both style and
content. The authors of this publication include not only the co-investigators of the Michigan and
Indiana study, but also the authors that contributed in substantial ways to this version.
Faber-Langendoen, D., J. Rocchio, S. Thomas, M. Kost, C. Hedge, B. Nichols, K. Walz, G. Kittel, S.
Menard, J. Drake, and E. Muldavin. 2012b. Assessment of wetland ecosystem condition across
landscape regions: A multi-metric approach. Part B. Ecological Integrity Assessment protocols for
rapid field methods (12). EPA/600/R-12/021b. U.S. Environmental Protection Agency Office of
Research and Development, Washington, DC.
117
-------
Appendix 4. LEVEL 3 PROTOCOLS
OVERVIEW
The intent of intensive methods for evaluating ecological-integrity is to develop data that are
rigorously collected, often with an explicit sampling design, to provide better opportunities to
assess trends in ecological integrity over time. Because of their cost and complexity, level 3
methods are often closely evaluated to ensure that they address key decision-making goals,
whether for restoration, mitigation, conservation planning, or other ecosystem management goals.
They are often highly structured methods, with detailed protocols that ensure a consistent,
systematic, and repeatable method (Sutula et al. 2006). The level of intensity required of level 3
methods typically means that they are used in conjunction with level 1 and 2 methods to increase
spatial representation and maintain affordability.
As with other levels, metrics that are chosen should be informative about integrity or sustainability of
major ecological factor or key ecological attributes and to associated stressors. Stressor tests can be
conducted by assessing how metrics respond to a gradient of stressors levels (Rocchio 2007, Jacobs et
al. 2010, Faber-Langendoen et al. 2011).
Level 3 metrics, more so than level 1 and 2, allow for greater specification by ecosystem type. The
detailed measures may allow for greater sensitivities in differences among ecosystems in terms of
ecological processes, structure or composition. Some intensive assessments have focused on one major
factor, that of vegetation. As with aquatic IBI methods (Karr and others), the approach has been to
focus on the biota, specifically vegetation, to develop a Vegetation Index of Biotic Integrity (VIBI( (Mack
2007). Quantitative vegetation sampling methods are well developed and relatively easy to implement
in the field, sampling is cost-effective and the data sets acquired from such sampling have multiple uses
including IBI development, setting mitigation wetland performance standards and supporting wetland
permit program decision-making (Fennessy et al. 2002). Level 3 assessments can also be expanded to
include soils and hydrology indicators.
118
-------
LIST OF LEVEL 3 METRICS
At this time, intensive metrics are still under development. We applied only one Level 3 metric to our
study.
FLORISTIC QUALITY ASSESSMENT (MEAN CN) MICHIGAN AND INDIANA
Definition: The mean conservatism of all the native species growing in the wetland.
Background: This metric is one aspect of the vegetation condition of specific occurrences of wetland
or terrestrial ecological systems. There are a variety of indices available to assess floristic quality, of
which Mean Cn is one. It is the average coefficient of conservatism across all native species for a give
sample unit or site.
The concept of species conservatism is the foundation of the Floristic Quality Assessment (FQA)
approach to monitoring and assessing ecological communities (Rocchio 2007). The core of the FQA
method is the use of "coefficients of conservatism" (C value), which are assigned to all native species in
a flora following the methods described by Swink and Wilhelm (1994) and Wilhelm and Masters
(1996). C values range from 0 to 10 and represent an estimated probability that a plant is likely to
occur in a landscape relatively unaltered from natural or historical range of variation (sometimes using
pre-European settlement conditions as the reference). A C value of 10 is assigned to species which are
obligate to high-quality natural areas and can't tolerate any habitat degradation whereas a 0 is assigned
to species with a wide tolerance to human disturbance. The proportion of conservative plants in a plant
community provides a powerful and relatively easy assessment of the integrity of both biotic and
abiotic processes and as such is indicative of the ecological integrity of a site (Wilhelm and Ladd 1988).
The mean Cn is the average C value across all native species. Mean Caii, which includes both native and
non-native species, where all non-natives are given a value of 0, is sometimes encouraged as providing
a more realistic account of the integrity of the vegetation within a site (Taft et al. 2006).
A Floristic Quality Index (FQI) can be derived from the C values (Swink and Wilhem 1994, Lopez and
Fennessy 2002). After each species has been assigned a C value, the average C value (mean C) of all
native species can be multiplied by the square root of site or total plot (or native) richness (Vs) to
produce the Floristic Quality Index (FQI) index, (also called the Floristic Quality Assessment Index, or
FQAI). Larger areas will typically support more species than smaller areas and since there may be cases
when a large and a small area share the same C value, accounting for species richness by multiplying it
with the C value adds a discriminating factor to the floristic quality assessment (Taft et al. 1997). Area
is not the only factor affecting species richness, as habitat heterogeneity and the presence of
anthropogenic patches can have an impact on richness, regardless of size (Wilhelm and Masters 1996).
Thus, to limit the influence of area alone on the index, the square root of species richness is used (Swink
and Wilhelm 1994; Taft et al. 1997). Still, interpretation of the index is more straightforward if a fixed
area is used, as species-area relationships can be directly interpreted, along with the index.
The index can be calculated using only native species, all species, species by cover, and other
permutations (see Table 2 in Rocchio 2), including the Adjusted Floristic Quality Index, which
eliminates the sensitivity of the index to species richness (Miller and Wardrop 2006). There is a FQA
version that relies on other field measures (wetland affinity status, exotic/native, and species richness)
that are more widely available than are C values (Ervin et al. 2006)
119
-------
Here we use only native species, presence absence.
Metric Type: Condition
Tier: 3 (intensive field method)
Rationale for Selection of the Variable: Plants grow in habitats to which they are adapted, including
biotic and abiotic fluctuations associated with that habitat (Wilhelm and Masters 1995). However,
when disturbances to that habitat exceed the natural range of variation (e.g. many human-induced
disturbances), only those plants with wide ecological tolerance will survive and conservative species
(those with strong fidelity to habitat integrity) will decline or disappear according to the degree of
human disturbance (Wilhelm and Master 1995, Wilhelm pers. comm. 2005).
Measurement Protocol: Species presence/absence data need to be collected from the wetland.
Although plot-based or area-based measurements are preferred, depending on time and financial
constraints, we measured this metric using a fixed area method, that of 1000 m2 described by Peet et
al. (1998). This method uses a 20 x 50 m plot which is typically established in a 2 x 5 arrangement of
10 x 10 m modules, and provides a standard 0.1 ha sample area, a widely used standard for assessing
species richness. However, the array of modules can be rearranged or reduced to meet site conditions
(e.g. 1x5 for linear areas or 2 x 2 for small, circular sites). Species presence and cover were recorded
in each of four modules. If time permits, the reset of the 50 x 20 m area can be surveyed for additional
species to obtain a 0.1 ha sample. The method is suitable for most types of vegetation, provides
information on species composition across spatial scales, is flexible in intensity and effort, and
compatible with data from other sampling methods (Peet et al. 1998, Mack 2004).
The metric is calculated by referencing only native species C value from a given state FQA Database,
summing the C value, and dividing by the total number of native species (Mean C). The Mean Cn is then
used to determine the metric status in the scorecard. The metric can also be calculated using all
species.
Metric Rating: Specify the narrative and numerical ratings for the metric, from Excellent to Poor (see
Master Table of Metrics and Ratings).
COEFFICIENT OF
CONSERVATISM
All wetlands
Metric Rating
Excellent
Cn = 6.0
Good
Cn 5.0 -5.9
Fair
Cn 3.5-5.0
Poor
Cn<3.5
Data: FQA methods have been developed and successfully tested in Michigan (Herman et al. 1996), and
Indiana (Coffee Creek Watershed Conservancy 2001).
Scaling Rationale: It is recommended that mean C and FQA index scores only be compared between
similar plant community or ecological system types (Rocchio 2007, Bowles and Jones 2006). In the
120
-------
Midwest, field studies using FQA have determined that a site with a Mean C of 3.0 or less is unlikely to
achieve higher C values (Wilhelm and Masters 1995). In other words, those sites have been disturbed
to the degree that conservative species are no longer able to survive and or compete with the less
conservative species as a result of the changes to the soil and or hydrological processes on site. Sites
with a Mean C of 3.5 or higher are considered to have at least marginal quality or integrity thus this
value was used as the Minimum Integrity Threshold (between Fair and Poor) (Wilhelm and Masters
1995). The threshold between Excellent, Fair and Good was assigned based on best scientific judgment
upon reviewing the FQA literature. However, mean C (and FQI) may not be a sensitive metric to detect
differences between Excellent and Good. For example, Bowles and Jones (2006) found that A and B
ranked dry to mesic prairies could be not discriminated based on mean C (or on FQI). The minimum C
value of 3.5 requires greater testing, since Rocchio (2007) found that heavily impacted sites still had
values above 5.5.
In central Pennsylvania, Miller and Wardrop (2006) found that for headwater wetlands, mean C n for
low impacted sites ranged from 4.55 to 6.13 (mean = 5.48 +. 0.46 S.D.), for moderately impacted sites
from 2.87 to 5.27 (mean = 4.17 + 0.74) and for highly impacted sites, from 2.0 to 4.78 (mean = 3.37 +
0.25). In the prairie pothole region, DeKeyser etal. (2003) found mean C n for low impacted sites >
4.01, for moderately impacted sites from 3.16-4.00, and for highly impacted sites, from 0.0 -3.15.
In West Virginia, Byers (pers. comm. 2007) found that the following C values (though all exotics were
assigned a value of "0", rather than being dropped from the calculations, which reduces the mean C of C
lower than if just natives are used). Data on the mean Coefficient of Conservation for 315 palustrine
plots throughout West Virginia for which EO Rank values have been assigned (EO Rank from Natural
Heritage Methodology):
Table 1. Coefficient of Conservatism Values (Caii) for wetland plots in West Virginia (Byers pers. comm.
2007)
Rating
Excellent
Good
Fair
Condition
Ranking
A
AB
B
BC
C
CD
MeanC
5.7
5.7
5.6
4.7
5.0
3.8
Number of
Plots
52
24
194
11
31
3
In Colorado, mean C n for highly impacted wetland sites had a mean approximately = 5.6, and for
reference, impacted sites, a mean of approximately = 6.7. The effectiveness of mean C n was best in The
effectiveness of Mean C (natives) for each ecological system type in Colorado Rocky Mountain wetlands
was very strong for fens, riparian shrublands, and slope wet meadows, though variability of the index
121
-------
for fens and riparian shrublands increased substantially as human disturbance increased. The index
was weakly effective in detecting human disturbance in extremely rich fens, and showed no utility for
riverine wet meadows.
Confidence that reasonable logic and/or data support the index: High
References
Andreas, B.K. and R.W. Lichvar. 1995. Floristic index for establishing assessment standards: A case
study for northern Ohio. Technical Report WRP-DE-8, U.S. Army Corps of Engineer Waterways
Experiment Station, Vicksburg, MS.
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid Bioassessment Protocols for Use
in Streams and Rivers: Periphyton, Benthic Macroinvertebrates and Fish, Second Edition. EPA
841-B-99-002. U.S. Environmental Protection Agency; Office of Water; Washington, D.C.
Bowles, M. and M. Jones. 2006. Testing the Efficacy of Species Richness and Floristic Quality Assessment
of Quality, Temporal Change, and Fire Effects in Tallgrass Prairie Natural Areas. Natural Areas
Journal 26(1): 17-30.
Coffee Creek Watershed Conservancy. 2001. 2001 Monitoring reports.
http://www.coffeecreekwc.org/ccwc/ccwcmission/monitoring reports.htm Coffee Creek
Watershed Conservancy, Chesterton, IN.
Dukes, J.S., and H.A. Mooney. 1999. Does global change increase the success of biological invaders?
Trends in Evolution and Ecology 14: 135-139.
Ervin, G.N. B. D. Herman, J. T. Bried, and D. C. Holly. 2006. Evaluating non-native species and wetland
indicator status as components of wetlands floristic assessment. Wetlands 26:1114-1129.
Fennessy, M. S., J. J. Mack, A. Rokosch, M. Knapp, and M. Micacchion. 2004. Integrated Wetland
Assessment Program. Part 5: Biogeochemical and Hydrological Investigations of Natural and
Mitigation Wetlands. Ohio EPA Technical Report WET/2004-5. Ohio Environmental Protection
Agency, Wetland Ecology Group, Division of Surface Water, Columbus, Ohio.
Herman, K.D., L.A. Masters, M.R. Penskar, A.A. Reznicek, G.S. Wilhelm, and W.W. Brodowicz. 1996.
Floristic quality assessment with wetland categories and computer application programs for
the State of Michigan. Michigan Department of Natural Resources, Wildlife Division, Natural
Heritage Program. In partnership with U.S. Department of Agriculture Natural Resources
Conservation Service, Rose Lake Plant Materials Center, Michigan
Ladd, D. 1993. The Missouri floristic quality assessment system. The Nature Conservancy, St. Louis,
MO.
Lopez, R.D. and M.S. Fennessy. 2002. Testing the Floristic Quality Assessment Index as an Indicator of
Wetland Condition. Ecological Applications 12(2): 487-497.
Miller, S.J. and D.H. Wardrop. 2006. Adapting the floristic quality assessment index to indicate
anthropogenic disturbance in central Pennsylvania wetlands. Ecological Indicators 6(2): 313-
326.
122
-------
Neckles, H.A., A. T. Gilbert, G. R. Guntenspergen, N. P. Danz, T. Hollenhorst, A. Little, J. Olker. 2007 (in prep).
Protocol for Monitoring Freshwater Wetlands in National Parks of the Northeast Temperate
Network. Prepared for the National Park Service. U.S. Department of the Interior. U.S. Geological
Survey, Reston, VA.
Neue, H.U. 1984. Organic Matter Dynamics in Wetland Soils. Wetland Soils: Characterization,
Classification, and Utilization. International Rice Research Institute. Manila, Philippines.
Northern Great Plains Floristic Quality Assessment Panel. 2001. Floristic quality assessment for plant
communities of North Dakota, South Dakota (excluding the Black Hills), and adjacent grasslands.
Jamestown, ND: Northern Prairie Wildlife Research Center Home Page.
http://www.npwrc.usgs.gov/resource/2001/fqa/fqa.htm
NRCS [Natural Resources Conservation Service]. 2001. Rangeland Soil Quality - Compaction. Soil
Quality Information Sheet, Rangeland Sheet 4. U.S. Department of Agriculture, Natural
Resources Conservation Service. Accessed online at:
http://soils.usda.gov/sqi/publications/sqis.html
Oldham, M.J., W.D. Bakowsky, and D.A. Sutherland. 1995. Floristic quality assessment system for
southern Ontario. Natural Heritage Information Centre, Ontario Ministry of Natural Resources,
Peterborough, Ontario.
Peet, R. K., T. R. Wentworth, and P. S. White. 1998. A flexible, multipurpose method for recording
vegetation composition and structure. Castanea 63: 262-274.
Pimentel, D., S. McNair, J. Janecka, J. Wightman, C. Simmonds, C. O'Connell, E. Wong, L. Russell, J. Zern, T.
Aquino, and T. Tsomondo. 2001. Economic and environmental threats of alien plant, animal,
and microbe invasions. Agriculture, Ecosystems and Environment 84:1-20.
Prichard, D., J. Anderson, C. Correll, J. Fogg, K. Gebhardt, R. Krapf, S. Leonard, B. Mitchell, and J. Stsitets.
1998. Riparian Area Management: A User's Guide to Assessing Proper Functioning Condition
and the Supporting Science for Lotic Areas. Technical Reference 1737-15. Bureau of Land
Management, U.S. Department of Interior, Denver, CO.
Richter, B.D., J.V. Baumgartner, R. Wigington, and D.P. Braun. 1997. How much water does a river need?
Freshwater Biology 37:231-249.
Rocchio, J. 2006. Vegetation Index of Biotic Integrity for Southern Rocky Mountain Fens, Wet Meadows,
and Riparian Shrublands: Phase 1 Final Report Unpublished report prepared for Colorado
Department of Natural Resources, and U.S. Environmental Protection Agency, Region VIII.
Colorado Natural Heritage Program, Colorado State University, Fort Collins, CO.
Rocchio, J. 2007. Floristic Quality Assessment Indices of Colorado Plant Communities. Unpublished
report prepared for Colorado Department of Natural Resources, and U.S. Environmental
Protection Agency, Region VIII. Colorado Natural Heritage Program, Colorado State University,
Fort Collins, CO.
Swink F. and G. Wilhelm. 1979. Plants of the Chicago Region. Revised and expanded edition with keys.
The Morton Aboretum, Lisle, IL.
Swink F. and G. Wilhelm. 1994. Plants of the Chicago Region. 4th Edition. Morton Arboretum, Lisle, IL.
123
-------
Taft, J.B., G.S. Wilhelm, D.M. Ladd, and L.A. Masters. 1997. Floristic Quality Assessment for Vegetation in
Illinois, A Method for Assessing Vegetation Integrity. Erigenia, Number 15, November 1997. The
Illinois Native Plant Society Journal.
Taft, J.B., C. Hauser, and K.R. Robertson. 2006. Estimating floristic integrity in tallgrass prairie.
Biological Conservation 131:42-51.
Wilhelm, G.S. and L.A. Masters. 1995. Floristic Quality Assessment in the Chicago Region. The Morton
Arboretum, Lisle, IL.
Yorks, T.E., and S. Dabydeen. 1998. Modification of the Whittaker sampling technique to assess plant
diversity in forested natural areas. Natural Areas Journal 18:185-189.
124
-------
Appendix 5. EXAMPLES OF METRICS & SCO RE CARDS
LEVEL 1 METRICS
Table Al. Great Basin Pinyon-Juniper Woodland EIA Scorecard (from NatureServe 2011)
Indicator
Justification
Rating
Sustainable
Transitioning
Degraded
Index
Score
Rank Factor: LANDSCAPE CONTEXT
Key Ecological Attribute: Landscape Connectivity
Connectivity
predicted by
Circuitscape
Intact natural conditions support
physical and biological dynamics
occurring across diverse
environmental conditions
Connectivity is moderate to
high and adequate to sustain
most CEs. Connectivity index
is>0.6
Connectivity is moderate to
low and will not some sustain
CEs. Connectivity index is 0.6-
0.2
Connectivity is low and will
not sustain many CEs.
Connectivity index is <0.2
0.73
Key Ecological Attribute: Landscape Condition
Landscape
Condition
Model Index
Land use impacts vary in their
intensity, affecting ecological
dynamics that support ecological
systems.
Cumulative level of impacts is
sustainable. Landscape
Condition Model Index is > 0.8
Cumulative level of impacts is
transitioning system between
a sustainable and degraded
state. Landscape Condition
Model Index is 0.8 -0.5
Cumulative level of impacts
has degraded system.
Landscape Condition Model
Index is< 0.5
0.88
Rank Factor: CONDITION
Key Ecological Attribute: Fire Regime
SCLASS
Departure
Mixed of age classes among
patches of the system is result of
disturbance regime. Departure
from mixture predicted under NRV
indicates uncharacteristic
disturbance regime and declining
integrity.
Mixed of age classes indicates
system is functioning inside or
near NRV. System is in a
sustainable state. Departure is
<20%. SCLASS Departure
Index is > 0.8
Mixed of age classes indicates
system is functioning near, but
outside NRV. System is
transitioning to degraded
state. Departure is 20 -50%.
SCLASS Departure Index is 0.8
-0.5
Mixed of age classes indicates
system is functioning well
outside NRV. System is
degraded. Departure is > 50%.
SCLASS Departure Index is <
0.5
0.50
Key Ecological Attribute: Native Species Composition
Invasive
Annual
Cover
Invasive annual vegetation
displaces natural composition and
provides fine fuels that
significantly increase spread of
catastrophic fire.
System is sustainable with low
cover of invasive annual
vegetation. Mean cover of
annuals is <5%. Invasive
Annual Cover Index is >0.8.
System is transitioning to
degraded state by abundant
invasive annual vegetation.
Mean cover of annuals is 5-
10%. Invasive Annual Cover
Index is 0.8-0. 5.
System is degraded by
abundant invasive annual
vegetation. Mean cover of
annuals is >15%. Invasive
Annual Cover Index is <0.5)
0.40
Rank Factor: SIZE
Key Ecological Attribute: Relative Extent
Change in
Extent
Indicates the proportion of change
due to conversion to other land
cover or land use, decreasing
provision of ecological services
provided previously.
Site is at or minimally is only
modestly changed from its
original natural extent (80-
100% remains) Change in
Extent Index is > 0.8.
Occurrence is substantially
changed from its original
natural extent (50-80%
remains). Change in Extent
Index is 0.8-0. 5
Occurrence is severely
changed from its original
natural extent (<50%
remains). Change in Extent
Index is < 0.5.
0.90
Overall Ecological Integrity Rank
(3.41/5 = 0.68) Mean Index Score
0.68
125
-------
LEVEL 2 METRICS
Adapted from Faber-Langendoen et al. 2008 (TablelOa).
Overview
*
Major
Ecological
Attribute/
Factor
LANDSCAPE
CON-EX-
SIZE
VEGE~A ION
HYDROLOGY
SOILS
Metric Name
La ndseaoe Co nneetvty
Land Use Index
Bjffe' Index
Paten S ze Co nd ton*
Paten S :zs ;.ha}
V e g e ta to n S t" uct ure
Vegetaton 'Co nposton
Re atve ~ ota Cover of Naive
P'antSpec-es
Invasve Exotc P-ant Spaces
Wate'Soj'oa
Hyd'ope^od
Hyd "0 og c -Con neotvty
Physca Patch ~yoes
So .' S j -face Co nd to n
Metrics Definition
A Teasjre of tne oereentof unf'agnented andscape w.tnrn 500 ™\ rad us.
A "••5a = jre ofine ntenstyof nj^ando"-, naied and jseswin naspeefc andscaaa a-sa^sjcn asa
caicn^ant, f*o"^tia -^eT.s'of •neocc.j'*ence. Eacn and jsetyoeoccj^ng nine andscaoaa'eas
a=sgnad a coeffcent -an^g ngf-3mO.Ota 1.0 ndcatng is ra atve ^-Da.;4. totnata"get sys'iS™1..
An ndexof'.neove'a a-eaandcond tDioflneojffer "•"•^sd atey sj^Djndngtnewet and, jsngS
"^vin s Dsac-o "acton; , and 5 jffer
Condt Dn, vVei and P jffar= a -a vegetated, n a: Jra Jnon-antn-D&Dgenc, a-aasinatsj^ojnda '.'.•aland.
A m",-aa = jre pf •tnecjrrem s zs of tne wetand ;'na; 'e aive:p tne p-g na natj'a =ze. Assessed py
dvd ng tne oasiesJ "-'as sfn stages za pyej^ant 33=3 jiesze.^Jt D ed oy 1M.
A "vaa = Jre Df tnecj"ent 5ze fna.. of tne D-sc-j^enjs o' stand. Assessed i-e-atve to refe'enos stands of a
rype.g ooa y. |
An assessment Dftneo-vera sfjclj'a 'M^pexlyoftnevegetBton ayers, .ne.jd.ng opesenc€;of njt.&e
st ata. age andsfjctj'a c-D^oextyofcanosy ayer, 3ndev"derEaQfdse3sear™ieiTta"V.
An assessr^ent oftne ove'a specescp^pos: Dnan>d d versty, nc-jdngpy aye-, andevdenceof
soecfcsoetesd se sees o ' "io la ty.
A "'•eas jra pf tne 'a at v'e oa-oa nt OP v-e- of a D ant spec-estnatare natve to tne "sgon. "" ypca y
"Tsasj-ed pye=tmatngto:a aoso jtecove-andsjDt.raavngtota'exot:csoee;es cover.
A "Tpeasu re of tne pe roent cove' of a set of exotc p a nt spaces that a re oo nsdered nvasve .
An assesses .nt of tne extent, d Jrat:on,andfreqjeney of saturated or ponded eondlons wl.n-na wetand,
as affected py the ends of d . recS .n p jls of water .nto , o p a ny d verso ns of wate " away f ro ™i, tne weta nd .
An assessment of tne c.naracters.tef'eqjercryar>d dd'atonof n jndatono'satj'atonof a wetanddjmg
a typca year.
An assess'Tentof tne ao lyof the wale .'to fow nto op out of the wetand.o'to ^njf>datfiad.aessntare^as..
A chectst of the nj^(De'rofd;fferantpnysea suntacesorfeatJ'esthatnTayppovde na&tatfop SDee.es
An assessment of so: surfaced stu^Qances [e.g. pa re so , t'ac<=J.
126
-------
LEVEL 3 METRICS
(from Tierney et al. 2009)
Table 2. Metrics and ratings for evaluating the ecological integrity of forest ecosystems
Metric type
Landscape
structure
Metric
Forest patch size
Anthropogenic land use
Good
>50ha
^_ 70% of stands are
late-successional
Rating
Caution
10-50 ha
10-40%
S/gnrficant Concern
40%
<70% of stands are late-successional in northern
hardwood, hemlock— hardwood, or upland spruce-
Structure
Stand structural class
Snag abundance
> 30% of stands are
late-successional
>. 25% of stands are
late-successional
>IO% of standing trees
are snags and > 10% of
med—Ig trees are snags
Coarse woody debris volume > 15% live tree volume
hardwood forest
< 30% of stands are late-successional in lowland
spruce—hardwood forest
<25% of stands are late-successional
for other forest systems
< 10% of standing trees
are snags or < 10% of < 5 med—Ig snags/ha
med-lg trees are snags
5—15% live tree volume < 5% live tree volume
Tree regeneration
Tree condition
Seed ling ratio >.0
Foliage problem < 10%
and no priority 1 or 2 pests
Seedling ratio <0
Foliage problem 1 0-50%
or priority 2 pest
Stocking index outside
acceptable range
Foliage problem >50% or
priority 1 pest
Composition
Biotic homogenization
Indicator species —
invasive exotic plants
No change
No key invasive exotic
plant species on most
plots
Increasing homogenization
One to three key species Four or more key species
per plot per plot
Function
Indicator species —
deer browse
Tree growth and mortality
rates
Soil chemistry — acid stress
Soil chemistry — nitrogen
saturation
No decrease in frequency
of most browse-sensitive
species
Growth ^ 60% mean
and mortal ityi 1.6%
Soil Ca:AI ratio >4
Soil ON ratio >2S
Decrease in frequency of Decrease in frequency of
most browsed species or most browsed species and
increase in frequency of increase in frequency of
browse-avoided species browse-avoided species
Growth <60% mean or mortality >l.6%
Soil Ca:AI ratio 1-4 Soil Ca:AI ratio < 1
Soil C:N ratio 20-25 Soil C:N ratio <20
Notes: MEfrd-ig trees are J^ 30 cm diameter-at-breast-hejghLTpee regeneranon stocking index vart&s by pirk. Priorffy I pests are Asian longhorned beetle, emerald ash borer.
and sudden oak death. Priority 2 pests are hemlock wooly adelgid. balsam wool? adelgid beech bark disease, and butternut canker. See cext for more deiai It.
127
-------
LEVEL 2 SCORECARD
(from Faber-Langendoen et al. 2011)
RANK FACTOR
MAJOR ECOLOGICAL FACTOR
/ Metric
ECOLOGICAL INTEGRITY
LANDSCAPE CONTEXT
LANDSCAPE
Connectivity
Land Use Index
BUFFER
Buffer Index
SIZE
SIZE
Relative Patch Size (ha)
Absolute Patch Size
CONDITION
VEGETATION
Vegetation Structure
Regeneration (woody)
Native Plants - Cover
Invasive Exotic Plants - Cover
Increasers - Cover
Vegetation Composition
HYDROLOGY
Water Source
Hydroperiod
Hydrologic Connectivity
SOIL
Physical Patch Types
Soil Disturbance
Rating
B
B
B
A
B
B
B
A
A
B
A
B
B
C
C
B
C
B
B
C
C
C
B
B
B
B
For each metric, a letter rating is assigned based on field or remote sensing data. Points are
assigned as follows: A = 5, B = 3.75 C = 2.5, D = 1.25. The rating is converted to a point value,
which is multiplied by the weight to get a metric score. Scores are summed within rank factors,
then divided by the summed weight to get a weighted average rank factor score and letter rank.
The final ecological integrity index and grade is based on first summing the three weighted
scores.
128
-------
LEVEL 3 SCORECARD
(from Tierney et al. 2009)
Metric type
Landscape
structure
Structure
Metric
Forest patch size
Anthropogenic landuse
Stand structural class
Snag abundance
CWD volume
Rating
Tree growth and mortality rates
Function Soil chemistry - acid stress
Soil chemistry - nitrogen saturation
Tree regeneration TBD
Tree condition
Composition Biotic homogenization TBD
Indicator species - invasive exotic plants
Indicator species - deer browse TBD
TBD
Ratings
Good
Caution
Significant concern
129
-------
LEVEL 3 METRIC SUMMARY BY MANAGEMENT UNIT
(adapted from NETN). SE = standard error.
Structure: Snag A bun dance
Abundance of snags by park or habitat
>= 10° o standing trees ancl -= 10° o med-lg trees are siiags
; 10° o standing trees or large snags under-represented
Significant Concern < 5 med-lg snags per ha
Park
% standing (SE)
% >= 30 cm dbh
N >= 30 cm dbh
ME]
AC AD MABI
10(2)
8(4)
2(2)
References
Faber-Langendoen, D., G. Kudray, C. Nordman, L Sneddon, L Vance, E. Byers, J. Rocchio, S. Gawler, G.
Kittel, S. Menard, P. Comer, E. Muldavin, M. Schafale, T. Foti, C. Josse, J. Christy. 2008. Ecological
Performance Standards for Wetland Mitigation based on Ecological Integrity Assessments.
NatureServe, Arlington, VA. + Appendices.
NatureServe. 2011. Central Basin and Range. Rapid Ecoregional Assessments, Final Memorandum 1-3-C
(March 4, 2011). Department of the Interior, Bureau of Land Management, Rapid Ecoregional
Assessments, Denver, Colorado. NatureServe, Washington DC.
Tierney, G.L, D. Faber-Langendoen, B.R. Mitchell. G. Shriver, J. Gibbs. 2009. Monitoring and evaluating
the ecological integrity of forest ecosystems. Frontiers in Ecology and the Environment 7: 308-316.
130
------- |