EPA
U.S. Environmental
Protection Ag»ncy
Washington, DC
EPA-SAB-EPEC-91-003
003
Report of The
Ecbregions
Subcommittee of The
Ecological Processes and
Effects Committee
Evaluation of The
Ecoregion Concept
A SCIENCE ADVISORY BOARD REPORT
CD '
~" »'"
f— •'
January 1991
EPA Headquarters Library
-------
,
'?•> -,>./•>.
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
OFFCEOF
THEADMWBTRATOH
January 22, 1991 EPA-SAB-EPEC-91-003
The Honorable William Reilly
Administrator
U.S. Environmental Protection Agency
401 M. Street, S.W. •
Washington, D.C. 20460
Dear Mr. Reilly:;
Ecoregions Subcommittee of the Science Advisory
JBoard_ (SAB) has comple€e*d~i"ts review of. the Ecoregions Concept
that was developed by scientists from the Environmental Research
Laboratory at Corvallis, Oregon. Ecoregions are geographic areas
that have similar variations in selected environmental
characteristics. These .environmental characteristics (e.g.,
chemical hardness of water and precipitation) are important
determinants of the types of biota that can exist in an ecoregion
and the quality that the environmental resource can attain.
Several states have already used the Ecoregion Concept to
delineate areas for water quality management and pollution
control. In addition, several Federal resource agencies are
considering using the-Ecoregion Concept to manage land and forest
resources and other EPA programs are considering regulatory uses.
The Subcommittee met on April 16-18, 1990, to review and
evaluate the Ecoregion Concept and hear the experiences of three
states (Arkansas, Ohio, and Minnesota), the Nature Conservancy,
and Environment Canada. The Subcommittee was asked to address
the following charge:
a. Can the Ecoregion Concept produce defensible and
reproducible classifications for any size areas?
b. Is the biogeographic and ecological science embodied in
the ecoregion concept developed well enough for states
to use this concept in their water standards program?
c. What are some current and/or future applications?
d. What .research is needed? ...
Priiatd on Rtcyeled Paper
-------
The Subcommittee believes that the Ecoregion concept is a
defensible classification technique for large areas (covering one
or more states) that is superior to the classification methods
that are currently used by most environmental managers. EPA is
to be commended for its accomplishments and leadership in
developing and promoting the use and understanding of the
concept. We have recently learned that EPA plans to cut
ecoregion research program in order to meet this years budget.
The Subcommittee believes that such a decision would be
unfortunate because it is one the few techniques available to
address ecological issues on a broad regional and global scale
that is needed to reduce ecological risks. In addition, many
states that plan to use this concept need technical support that
can be sustained through an active research program.
Several states have demonstrated that the Ecoregions Concept
also has application as a water quality management tool within
states. The Subcommittee believes that states need assistance
from EPA in applying the Ecoregion Concept to the development of
water quality criteria and standards to promote reproducible
results. In a few cases, states have developed biological '
criteria for those areas. The Subcommittee did not review the
basis for biological criteria for water quality; rather it
recommends that a separate review be made of the concepts and
technical guidance for biological criteria. In addition, more
technical guidance and assistance is needed to help users address
complex problems such as rivers that cross ecoregion boundaries.
The Subcommittee concluded that the Ecoregion Concept has
many other potentially useful applications. In addition to the
water quality management uses, land, wildlife, and timber
management have been aided by this regionalization approach which
helps to identify areas of similar ecological potential. Some
regulatory programs are considering using ecoregions to designate
areas for chemical releases. In the future, ecoregions may be
useful to develop the criteria for multimedia or cluster
regulatory approaches. Finally the Ecoregions Concept may have
application to identify areas for monitoring and assessment
activities, including the Environmental Monitoring and Assessment
Program (EMAP).
Further research and evaluation is needed on the Ecoregion
Concept which should include the following: a pilot test to
determine whether and to what degree ecoregions perform better
than other less sophisticated regionalizations, methods for
defining and locating boundaries, methods for selecting reference
sites, and formal procedures for delineating and subdividing
ecoregions. We recommend that much of the research on
ecoregions be coordinated with EMAP.
-------
Although we expect that the program office will address all
of the issues presented in this report, we particularly direct
your attention to the following ones:
- the lack of support for research or technical
assistance and specific written guidance that is essential
for states and other potential users.
-the potential applications of the ecoregion concept to many
resource management issues, including strategies for
reducing risk.
-the need for objective procedures to select reference
sites, define boundaries, and subdivide ecoregions.
The SAB appreciates the opportunity to.conduct this
scientific review and looks forward to receiving your response, to
the scientific advice transmitted herein.
Sincerely,
Dr. Rymond Loehr, Chairman
Executive Committee
Science Advisory Board
Dr. Kenneth Dickson, Chairman
Ecological Processes and
Effects Committee
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
NOTICE
This report has been written as a part of the activities of
the Science Advisory Board, a public advisory group providing
extramural scientific information and advice to the Administrator
and other officials of the Environmental Protection Agency. The
Board is structured to provide a balanced expert assessment of
scientific matters related to problems facing the Agency. This
report has not been reviewed for approval by the Agency; and,
hence, the contents of this report do not necessarily represent
the views and policies of the Environmental Protection Agency or
other agencies in Federal government. Mention of trade names or
commercial'products does not constitute a recommendation for use.
-------
ABSTRACT
This report presents the conclusions and recommendations of
the U.S. Environmental Protection Agency's Science Advisory Board
following a review of EPA research on the Ecoregion Concept and
presentation of the results of its application for water quality
management by three states. The Ecoregion Concept is method of
dividing large geographic areas in regions or subunits in which
the variability of selected ecological and physical
characteristics is less than that of the entire area. The
Ecoregion Concept, as published by EPA, is being used by states
for water quality management. The principal concerns of the
Subcommittee are that limited guidance and documentation is
available to users for defining and locating the boundaries and
establishing adequate reference sites and that informal methods
are used to subdivide areas. The Subcommittee endorsed the
concept but recommended that EPA renew and sustain its research
in critical areas, conduct a pilot project to compare the
effectiveness of Ecoregions with other regionalization
techniques, and develop a user guidance with case studies to
assist future applications.
Key Words: Ecoregions; Regionalization; Environmental
Management.
11
-------
r
U.S. EKVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD
BCOREGIONS SUBCOMMITTEE
ROSTER
CHAIRMAN
Dr. Kenneth Dickson
Director
Institute of Applied Sciences
North Texas state University
P.O. Box 13078
Denton, Texas 76202
MEMBERS/CONSULTANTS
Dr. Michael Goodchild
Co-Director
National Center for Geographic
Information and Analysis
Department of Geography
University of California
Santa Barbara, California 93106
Dr. Daniel Goodman
Montana State University
Department of Biology
Louis Hall
Bozeman, Montana 59717
Dr. J. Fraser Hart
Department of Geography
414 Social Science Building
267 19th Avenue South
University of Minnesota
Minneapolis, Minnesota 55455
Dr. Carolyn Hunsaker
Environmental Sciences Division
Oak Ridge National Laboratory
Oak Ridge, Tennessee 37831
iii
-------
Dr. Richard Kimerle
Monsanto Corporation
800 N. Lindbergh Boulevard
St. Louis, Missouri 63167-5842
Dr. John Neuhold
Department of Wildlife Sciences
College of Natural Resources
Utah State University
Logan, Utah 84322
Dr. William Platts
Don Chapman Consultants, Inc.
3180 Airport Way
Boise, Idaho 83705
Dr. Paul Risser
Vice President for Research
University of New Mexico
Albuquerque, New Mexico 87131
Dr. 6. Bruce Wiersma
Director, Center for Environmental
Monitoring and Assessment .
Idaho National Engineering Laboratory
ILF 33
EG&6 Idaho, Inc.
P.O. Box 1625
Idaho Palls, Idaho 83415
SCIENCE ADVISORY STAFF
Dr. Edward S. Bender
Designated Federal Official
U.S. Environmental Protection Agency
Science Advisory Board
401 M street, SW.
Washington, D.C. 20460
Mrs. Frances A. Dolby
Secretary to the Executive Secretary
IV
-------
TABLE OP CONTENTS
1.0 Executive Summary 1
2.0 Introduction •. 2
2.1 Charge to the Subcommittee 3
2.2 Subcommittee Review Procedures 3
3.0 Evaluation of the Ecoregion Concept 4
3.1 Validity of Ecoregions 4
3.1.1 Geographical Considerations ......... 5
3.1.1.1 Variables 5
3.1.1.2 Ecoregion Size ... 5
3.1.1.3 Number of Subdivisions ... 6
3.1.1.4 Contiguity 6
3.1.1.5 Hierarchy 6
3.1.1.6 Homogeneity. . 7
3.1.1.7 Unique Assignment 7
3.1.2 Ecological Considerations .... 7
3.2 Use of Ecoregions for Water Quality Management
8
3.2.1 Reference Sites 8
3.2.2 Resolution of Boundaries 9
3.2.3 Subdividing Ecoregions 10
3.2.4 Formalization 11
3.2.5 Qualitative versus Quantitative Methods . . . 11
3.2.6 Testing 12
3.2.7 State Water Quality Uses 13
3.3 Other Applications 14
4.0 Research Needs
5.0 Summary of Recommendations
6.0 Literature Cited
Appendix A. Examples of Testing Ecoregions . .
DRAFT EVALUATION OF ECOREGIONS
16
17
19
22
1.0 EXECUTIVE SUMMARY
EPA has conducted extensive research to develop Ecoregions
of the conterminous United States (Omernifc, 1987). Ecoregions
are derived qualitatively by geographers working with
environmental scientists and they show areas where several
-------
environmental variables are more similar within a region than•
between regions. Ecoregions have been used by several states,
with support from EPA, as a framework: for environmental quality
management and pollution control. EPA's research program on the
development of the Ecoregion Concept was completed in June, 1989.
The Ecoregions Subcommittee was asked to review the
Ecoregion Concept for scientific merit, adequacy for use by
states, other potential applications, and to identify research
needs. The review was conducted at Corvallis, Oregon in April,
1990 included briefings by the developers of the Ecoregion
Concept and presentations by several state users and a scientist
from Environment Canada. In support of the research and its
applications, EPA and collaborating scientists have published
more than twenty peer reviewed articles over a five year period
which were consulted as part of this review.
Overall, the Subcommittee finds that the Ecoregion Concept .
is defensible for classifying large (multi-state) areas when used
by skilled professionals. EPA is to be commended for its
accomplishments and leadership in developing and promoting the
use and understanding of the concept. The Subcommittee believes
that the Ecoregion Concept will lead to a method of subdividing
geographic areas that is superior to the methods that, are
currently used by most environmental managers. The Subcommittee
believes that it is unfortunate that after promoting the method
and eliciting interest among state and Federal users, EPA has
stopped its research on the concept and eliminated its efforts
toward technology transfer. The Subcommittee further notes that
continued research is in keeping with the new emphasis on
ecology, in general, in the "Reducing Risk ..." report of the
SAB. The Subcommittee recommends that EPA develop a pilot
project with the explicit aim of determining whether and to what
degree defined ecoregions perform better than other less
sophisticated regionalizations in ecological management.
Several states have used the Omernik Ecoregions as a basis
for dividing their states into water quality use areas and in a
few cases, they have developed biological criteria for those
subdivisions. While the state record of achievement is good, it
was developed under the tutelage of EPA, and several topics
require further research and testing: the definition and location
of boundaries, selection and number of reference sites, and
formal procedures for delineating and subdividing ecoregions.
Due to these research needs, the lack of quantitative methods for
testing regionalizations and limited user guidance, the
Subcommittee believes that a relatively high level of expertise
is required to produce defensible and reproducible subdivisions
within state areas. The Subcommittee recommends that EPA develop
guidance and analyze the results of past applications of
ecoregions to assist future use by the states. The Subcommittee
-------
did not review the basis for the states' biological criteria;
however, it recommends that a separate review be made of
biocriteria, perhaps after technical support documents become
available.
The Subcommittee concluded that the Ecoregion Concept has
many potential useful applications. In addition to the water
quality management uses, land and timber management have been
aided by this regionalization approach which helps to identify
areas of similar ecological potential. In some cases, ecoregion
boundaries may replace political boundaries that are used by some
regulatory programs to designate areas for restricted chemical
uses. In the future, the Ecoregion Concept may be used to
develop criteria for multimedia regulations. Finally the
Ecoregions Concept may -be used to identify areas for monitoring
and assessment activities, including the Environmental Monitoring
and Assessment Program (EMAP).
2.0 INTRODUCTION
EPA has conducted extensive research on the techniques for
identifying environmental regions that can be used for ecosystem
management. Most of the research on regionalization in EPA has
been conducted at the Office of Research and Development (ORD)
Environmental Research Laboratory (ERL) at Corvallis, Oregon. In
1987, EPA published maps in the peer reviewed literature of
Ecoregions for the Conterminous United States (Omernik, 1987).
EPA has worked with several states to apply the concept to the
water quality standards process (Gallant, et al., 1989)'. Even
though EPA's effort on the Ecoregion Concept ceased; however,
several states and EPA program offices are now investigating
further applications of the Ecoregion Concept. Recently, the
Office of Water (OW) established a policy (EPA, 1990) that
encourages states to develop biological criteria for water
quality protection. The Ecoregion Concept is being used by
states in the development of biological criteria.
The Science Advisory Board was aware of the widespread and
growing interest and potential applications of the Ecoregion
Concept. In December 1989, the SAB agreed to undertake this
review because the ecoregion concept is being used by other
Federal agencies, several states, and potentially could be used
by many others. Furthermore the concept is closely related to
the new and developing field of landscape ecology.
The review was assigned to the Ecological Processes and
Effects Committee (EPEC) by the Executive Committee of the SAB.
EPEC formed the Ecoregions Subcommittee to conduct the review.
The review was hosted by the U.S. EPA Environmental Research
Laboratory at Corvallis, Oregon on April 16-18, 1990.
-------
2.1
e to
The Subcommittee received an informal charge from the Office
of Environmental Processes and Effects Research that was modified
by the subcommittee at the review meeting to reflect the concerns
of the EPA researchers and the needs of the state users of the
Ecoregion Concept. The Subcommittee accepted the charge to
address the following questions:
a. Can the Ecoregion Concept be used to establish
defensible and reproducible classifications for any
size areas? What is the nature and extent of the
uncertainty associated with establishing boundaries?
b. Is the biogeographic and ecological science embodied in
the Ecoregion Concept developed well enough to justify
states using this concept in their water quality
standards program? Are there limitations or
qualifications which should be noted as part of such an
application?
c. What are some current and/or future applications which
one might envision for the Ecoregion Concept?
d. What research is needed to develop the concept further
and assure the validity of its results?
2.2 subcommittee Review Procedures.
The Agency provided the Subcommittee with a number of
publications and reports (see Literature Cited) which were
reviewed by the subcommittee members in preparation for the
evaluation. From the background material, the subcommittee
developed preliminary impressions and questions for follow-up at
the review. Nearly two days of presentations were made at the
review, including a report from Canada on their use of ecological
regions and the experience of three states that had used, the
Omernik (1987) ecoregions as a basis for defining use
attainability and biological criteria as part of their water
quality standards program. One of the States had used ecoregions
to classify lakes.
3.0 EVALUATION OP THE ECOREGION CONCEPT
The concept of ecoregions has continued to evolve from the
original term of Crowley (1967) and the first mapped
classification of ecological regions of the United States by
Bailey (1976) . The Ecoregion Concept is a special method of
regionalization for subdividing a geographic area into regions of
relative homogeneity in ecological systems or in relationships
between organisms and their environments. The resulting
-------
ecoregions can provide a valuable framework for environmental
resource managers and scientists to use for monitoring,
assessment, and control measures. This is unfortunate since the
development of Ecoregions Concept by the Corvallis ERL was done
with modest financial support from ORD and the Office of Water,
however, since June 1989 there has been no active research
program on the Ecoregions Concept. The development of the
Ecoregions Concept is an example of creative and proactive
scientific research in the Agency, overall, the subcommittee
finds that Scoregions and regionaliiatioa are valuable concepts
with many potentially useful applications for environmental
management, and further the subcommittee encourages the Agency to
resume development and testing of the concept to assess its
strengths and limitations for further applications.
EPA has used the regionalization process to develop several
different kinds of ecoregions, ranging from single variable maps
of alkalinity regions for the acid rain program to maps (Omernik,
1987) which describe areas of similar vegetative and hydrological
characteristics. The Omernik ecoregions were used to illustrate
applications by states and were the primary focus for this
review. However, the comments in this report can be broadly
applied to any form of regionalization, and ecoregions are not
restricted to those specifically developed by Omernik.
3.l Validity of the Beoreqion concept
The Subcommittee was asked to assess 1} if the Ecoregion
Concept could be used to establish defensible and reproducible
classifications for any size areas and 2) the uncertainty and
decisions associated with establishing ecoregion boundaries.
EPA must demonstrate not only that the Ecoregion Concept
is useful but that it outperforms other forms of regionalization
in its particular applications. This is a challenging task. It
is important to find out where ecoregions perform well and where
they do not, and it is worth expending significant resources on
this effort.
. The Ecoregion Concept is based on the premise, long a tenet
in the professions of geography and environmental sciences, that
the world becomes more understandable when its surface is divide 1
into logical units based on some process, feature or activity.
The variance of selected parameters within these units is
reduced. Local outliers are more easily discernible (if local
data, independent of the ecoregion map are available).
The Subcommittee considers classification and
regionalization (mapping) to be an evolving process.
Regionalization often depends on assumptions, judgments, opinicr •>
and some data and it is not easily quantified. The initial
regionalization schemes are developmental and need more testing.
-------
As experience is gained and. as different sets of regions are
evaluated with respect to different questions, particular schemes
should become more readily supported by facts and amenable to
quantification. .
3.1.1 Geographic Considerations.
Many professional judgments were made to develop the Omernik
(1987) ecoregions. These include judgments on the appropriate
variables, ecoregion size, the number of subdivisions,
contiguity, hierarchy, homogeneity and unique assignment. The
implications of these .judgement.for users of ecoregions are
detailed in the following paragraphs.
3.1.1.1 Variables. The Omernik ecoregions are based broadly
on four primary variables: land surface form, soils, natural
vegetation and land use. A continuum of possible
regionalizations can exist.- At one end are special-purpose
regionalizations based on single variables, e.g., soils, and
serving limited purposes. At the other-end is a regionalization
based on all climatic and geological variables which attempts to
serve all purposes. By using four primary variables and choosing
others in some situations, EPA made decisions which can' limit the
future applications and subdivisions of the regionalisation map.
such limitations or special precautions should be explained and
documented to guide further users of the map.
3.1.1.2 EcoreaionSize. Gallant et al. (1989) indicated
that there may be areas where the mapped numerical field data do
not correspond with the distribution of environmental features.
They suggest that it may be useful to map such areas.as separate
regions if they are of comparable size to other regions being
mapped. This advice is somewhat confusing, because the nature, of
geographic variation suggests that minimizing heterogeneity will
not lead to. regions of similar area. In fact quite the opposite
is true. Geography consists of large areas of uniformity, with
relatively small areas of complex variation, so a regionalization
based .on minimizing heterogeneity would lead to very high
variability in region size. Perhaps the authors should clarify
their definition of comparable size areas.
3.1.1.3 Number of subdivisions. It is possible to divide
the surface of the United States into any number of regions.
There is a simple monotonic relationship between the number of
regions and their homogeneity. More regions will always result.
in less heterogeneity or a more accurate representation of true
spatial distributions (Jensen et al., 1989) within each region.
For example, one might specify the number of regions indirectly
by deciding that a certain mean region area was desirable for
management purposes. The criteria need not be scientific. Th«
Subcommittee recommends that the criteria used to establish th«
-------
number of subdivisions be laid out as explicitly as possible.
The consensus of the numerical taxonomy literature (see for
example Sneath, P.H. and R.R. Sokal Numerical Taxonomy. San
Francisco, Freeman, 1973; Jardine, N. and R. Sibson Mathematical
Taxonomy New York, Wiley, 1971.) is that there is no
satisfactory, objective way of defining the number of regions or
clusters "naturally" present in data. Information is lost when
the specific attributes of a case are replaced by the attributes
of a cluster, and the information loss increases from zero, when
every case is its own cluster, to a maximum when all cases belong
to one cluster. The increase is monotonic on most measures of
information loss, even though measures can be devised for which
that is not true (e.g., the slope of the so-called "scree"
diagram).
3.1.1.4 Contiguity.. If places are classified solely on the
basis of the characteristics present, then when the classes are
mapped, the resulting regions may not be contiguous (singly
bounded). one class may consist of several islands, such as the
Western Forested Mountains Ecoregion (Omernik, 1987). It is
possible to require single boundedness, but in general 10 singly-
bounded regions will not be as homogeneous as 10 unconstrained
•classes. The decision as to when to require single-boundedness
is aa important part of the regionalisation process, which should
be explained in documentation supporting the regionalizatioh.
3.1.1.5 Hierarchy. Geography is characterized by the
expression "the closer you look, the more you see". It is
appealing to think that subregions are hierarchically nested
within one region. By subdividing regions into smaller and
smaller areas,, we .can capture a closer and closer approximation
to the truth. However, hierarchical nesting is not a natural
property of.regions, but rather an artifact of the
regionalization process. For example, on a broad scale of
regionalization, using data or maps with low resolution, data for
several parameters may be aggregated so that a boundary line may
represent a zone of transition for a single linear feature (e.g.,
elevation). At a finer scale of regionalization, data may be
aggregated for individual parameters and there will be more
subregions. And it is likely that the zone of transition between
regions will become one or more subregions. In such cases, two
or more regions may have areas that are common to a single
subregion. In principle, then, subdividing large regions does
not always allow us to locate boundaries of the subregions more
accurately. Therefore, imposing hierarchical nesting on the
different levels of regionalization may be useful from a
management perspective, but may be more likely to conflict with
the accurate representation of spatial variation. Thus one must
weigh the desirability of having nested regions against the need
for greater accuracy.
-------
3.1.1.6 Homogeneity. Although.it is natural to assume that
regions are defined by uniform characteristics, it is not
uncommon for a region to be defined as a mixture of substantially
different characteristics. For example, the "ridge and valley
province" is uniform in its intermingling of two very different
landforms. One of the decisions to be made in setting toe goals
of a regionaliiatioa activity is the degree to which homogeneity
vill. be required. This has implications for the definition of
archetypes or reference sites, since there may be no archetypical
location in a region that is defined by a mixture of
characteristics.
3.1.1.7 Unique Assignment. The regionalization for
Ecoregions assigns each location on the earth's surface to
exactly one class, and thus it has precise linear boundaries.
But for scientific purposes it is possible to work with other
models of regionalization. We might, for example, define a
number of reference sites across the United states typifying its
geographical variation, and then characterize each location of
interest by its similarity to the reference sites. Presumably
each place would be similar to more than one, but not to all
reference sites. From a scientific perspective this is perhaps a
more effective way of characterizing multivariate spatial
distributions than regionalization. While such an approach is
more difficult to map, modern spatial database technology makes
it feasible using digital information. The usefulness of one
approach over another for environmental management and ecological
research should be evaluated by a well-designed pilot test.
3.1.2 geological Considerations. «
As mentioned earlier, the process of ecoregion definition
must be an evolutionary one. Ecoregions can be distinguished by
measuring the variance of indicators of ecological condition or
health and designating reference sites in which those indicators
exhibit nominal conditions. The Environmental Monitoring and
Assessment Prograa (EMAP) vill be measuring a variety of-
ecological indicators across the nation and it could provide a
useful framework for testing and refining the Ecoregion Concept.
The attributes on which some ecoregions are determined may
change over time. The two most obvious changing conditions are
land-use and vegetative successional states, but climate may also
change. Thus, depending upon the scale of resolution, regional
delineation may also need to change. EPA should provide advice
to the users for monitoring to determine when the original
regionalization is no longer valid for its intended purpose and
how one should go about altering the regional!zation.
3.2 Use of Ecoregions for WaterQuality Management
-------
The Subcommittee was asked to evaluate whether the
biogeographical and ecological sciences embodied, in the Ecoregion
Concept ar« veil enough developed to justify states using this
concept in their water quality management program.
The Subcommittee responded to this request with positive
but reserved reactions. The effort is an innovative approach to
environmental quality management. It is potentially very useful
in recognizing deviation of ecosystems, from a regional norm, in
response to changing water, soil and air quality. It provides a
sounder (and much more equitable) basis than is currently used
(state boundaries) for establishing environmental quality
standards for defined ecosystems within an ecoregion context.
The Subcommittee have had reservations about the use of
ecoregions by states because additional guidance is needed to
assist them with selection of reference sites, resolution of
boundaries, and further subregionalization of the Omernik
ecoregions. While several states have successfully applied
ecoregions in their water quality standard programs, they have
had substantial assistance from EPA. Currently, it is unclear
what type of assistance will be available for state needs in the
future. Therefore, the subcommittee recommends that EPA develop
.guidance for states to use ecoregions and provide technical
assistance as veil.
3.2.1 Reference Sites. It is important to use regional
reference sites when the Ecoregion concept is applied to protect
ecosystems. The biological, physical and chemical
characteristics of these regional reference sites can be used to
establish the recovery potential for impacted aquatic systems in
the same ecoregion.
Users of the ecoregion concept are instructed to identify
"minimally impacted sites" (Hughes et al., 1986) as benchmark
conditions for a specific ecoregion. The Subcommittee strongly
recommends that EPA develop user guidance with criteria.for the
unbiased selection and appropriate number of regional reference
sites.
Currently, criteria to select regional reference sites may
differ depending on the application of the Ecoregion Concept.
For example, if the application is to develop biological criteria
for impacted stream segments in an Ecoregion, it may be necessary
to identify and monitor only minimally impacted reference sites.
However, if the application of the Ecoregion Concept is to assess
the efficacy of best management practices for non-point source
pollution control, then the regional reference sites should
represent sites that are impacted by a variety of non-point
source problems of different intensities. In this application,
reference sites provide estimates of biological, chemical and
physical attributes that may be expected using different degrees
-------
of best management practices.
A stochastic sampling approach may be used to select
reference sites. In this case using a set of criteria (see for
example Hughes et al, 1986) one would define the universe of
potential reference sites for a particular ecoregion, then use an
unbiased selection process to make the final selection (i.e.,
enumerate all the potential reference sites and then make random
choice of sites).
The number of reference sites is not only a function of the
statistical variance of the selected parameters, but also of the
size of the potential universe of such sites and of the funds
available.
3.2.2 Resolution of Boundaries.
For the Ecoregion Concept, boundaries take on a
critical importance because the interiors of ecoregions are
treated as homogeneous. However, with the exception of perhaps
political boundaries, boundaries of regions are fuzzy because
they represent a transition or a gradient of change between
regions. These gradients may differ in steepness.
Regionalization at the required level of resolution can provide
the basis for the extrapolation of results from one basin to
another, from one habitat to another, and from one biological
assemblage to another within a region. Thus, the research
questions are how to define the locations of boundaries, how to
represent them on useful maps, and how to interpret them in the
context of the attributes that define the ecoregion.
Ecoregions can be developed for various spatial scales and
can be narrowly or broadly focused depending on the desired
application. A regionalization scheme that is broad-based and
appropriate for a large variety of uses will most likely be based
on processes which control ecological conditions. For example,
the Omernik ecoregions represent the controlling processes of
geology and climate (physiography, soils, land use and potential
vegetation). The more variables or maps that support the
location of an ecoregion boundary, the more robust that boundary
should be. Omernik ecoregions were developed for aquatic
systems; they may be quite useful for terrestrial systems because
the same component data/maps would likely be used; however, the
exact location of some of the boundaries might be different for
terrestrial systems. Specialized ecoregions also have been
developed and used for acid rain research (alkalinity patterns)
and lake trophic state (phosphorus patterns).
Present ecoregions should continue to be subdivided by
higher resolution boundaries until maps are available at the
appropriate resource level for different applications. Once
10
-------
ecoregion boundaries are fine tuned to the level that they will
answer the general questions being asked, then the ecoregions
could be subdivided for specific purposes, e.g., to determine and
monitor proper Best Management Practices for livestock grazing,
logging and road construction, fisheries, and mining. Platts et
al. (submitted, 1990) examined the relationship between
classification and the stream flow requirements which are
necessary to maintain riverine-riparian habitat and valley bottom
type. They recommended a specific sequence of analysis to
classify land to the valley bottom type level of resolution.
The value and usefulness of an ecoregion map improves when
the decisions on geographic characteristics are documented. The
level of confidence for the location and width of a region's
boundary also needs to be documented in the future. The
components that played a primary role in boundary placement can
even be indicated (Clarke et al., submitted, 1990).
3.2.3 Subdividing Ecoreaions.
The most useful approach to ecological regions for a
national framework is a set of broad-based regions at different
spatial scales. An excellent example of such a national
:framework is Environment Canada's ecological regions (Wiken 1986)
which have seven levels of regionalization classes. These
ecological regions are still evolving and are being successfully
used for planning, assessment, and management of many resources
including wildlife, forests, wetlands, agriculture, as well as
for addressing issues such as acid deposition and climate change.
EPA's work with ecoregions provides regions appropriate at the
(1) national scale (7 classes, Omernik and Gallant 1989), (2)
regional scale (57 classes, Omernik 1986), and (3) regional/state
scale (Omernik and Gallant 1986, 1987a, 1987b, 1987c, 1988;
Omernik 1987a, 1987b, 1987c). The same regionalization process
could be used to develop even smaller ecoregions (higher
resolution) for other uses such as biological criteria.
It is clear that the Ecoregions Concept is more useful for
states if the national map of ecoregions is subdivided to reflect
the ecological systems that occur within states. At the state
level, more explicit decisions and criteria are required because
of the proximity to regulatory and planning decisions. Moreover.
there is some advantage for having uniform rules from state to
state. Therefore, the Subcommittee recommends that EPA develop a
process for creating subdivisions that is defensible in many If
not all states.
EPA should consider subdividing Ecoregion map so maps are
available at the land type and valley bottom type levels of
analysis. To date most time and effort has been spent working i-
the national level of ecoregion analysis. This fundamental level
had to be established, but nov effort should be directed to
11
-------
d«*on»trat« the •ff«otiv«n«a« of r«gionali«ation at tn« level
wh«r* state and local •an«g«a«at decisions are b«ing made.
3.2.4
Formal izat ion of the process for delineating ecoregions and
subregions offers the prospect of making the task of creating
regional maps more rapid and inexpensive. Regionalizations could
be more reproducible, more nearly uniform, and more objective.
Finally it could define general standards of performance to judge
specific regional izat ion applications.
In a formalized context, the quality of the product can be
described in terms of uncertainties in quantities, such as
location of boundaries and inhomogeneities between sites within a
given boundary, that are important to management and policy uses
of the product. A paper submitted by Clarke et al provides a
useful approach toward documenting the uncertainties associated
with some ecoregion boundaries. The tolerances for uncertainty
in the product. can be used to calculate the input data quality
which is required. Once the requirements for input data quality
are known, a Quality Assurance/ Quality Control program can be
developed. > - '
The regionalization process should • include a formal
objective protocol, with explicit QA/QC standards, and
regionalization maps should include quantitative statements on
their expected performance. Very little has been done, to date,
in the direction of accomplishing this. .
3.2;5 Qualitative versus Quantitative Methods.
*
The Subcommittee did not reach consensus with respect to the
status of currently available methodology for regionalization.
All of the Subcommittee members agreed that a formal, quantified
approach would be desirable in principle, but there was
uncertainty over whether such an approach could be implemented
without further, extensive work on methods development. Some
members felt that the present state of the relevant mathematical
and statistical science and computational technology is adequate
for immediate adoption of a formal quantified method; other
members were not so certain.
Th« Subcommittee wishes to reiterate that, notwithstanding
the absence of consensus on presently feasible methodology for
regionalisation, the committee was unanimous in its enthusiasm
for the Ecoregions Concept. While Omernik Ecoregions, as
presently implemented, 'rely extensively on informal, qualitative
"professional judgement", the concept still represents an advance
over other, more subjective, frameworks for deciding questions
such as regional standards for attainable environmental quality.
12
-------
The current practice is that delineation on maps is carried
out with some mixture of quantitative and qualitative techniques,
with their respective advantages and disadvantages. The
subcommittee recommends that future research should seek to
increase the degree of formality and quantification, by using
computational (including multivariate statistics and artificial
intelligence) techniques wherever possible, and by stating the
implicit • rules* for the remaining qualitative components as
explicitly as possible, so that the remaining margin of
subjectivity can be understood by the user. The statistical
properties, including reproducibility, of the boundaries
generated by the state-of-the-art regional izat ion process should
be quantified, at intervals, as the methods evolve.
3.2.6
Testing of regional izat ions presents several important
technical problems related to the difficulty of dealing with
statistical data in a spatial domain. These problems are evident
in several of the papers dealing with validation. An example of
validating a regional ization hypothesis was discussed in section
3.1.1.7. .The results of several applications of ecoregions are
tested and discussed further in appendix A.
While a program or method for quantitative delineation of
regions does not currently exist, it is likely that a few years'
effort could develop such a technique, given the modern
accomplishments in computer pattern recognition, computer image
enhancement, and spatial statistics. EPA should demonstrate
that "qualitative" methods are reproducible and attempt to apply
quantitative methods so that the performance of ecoregions
relative to other regional izat ion schemes can be evaluated.
Regionalization is a potentially powerful method, both for
science and management. It represents a way of viewing spatial
variation which is particularly valuable when land is classified
into discrete categories. For other types of applications it may
be less appropriate, and certainly should not be seen as' a
universal solution to spatial data management. For variables
measured on continuous scales, such as elevation or atmospheric
pressure, contours are a more efficient method of representation
involving much less information loss. However, the map of summer
total phosphorus in lakes in Minnesota, Wisconsin and Michigan
(Omernik et al., 1986) shows a continuous variable that is quite
homogeneous within some regions with abrupt changes between
regions. For example, the low phosphorus concentrations in lakes
of region 50-6 ( less than 5-14 micrograms/1) occur adjacent to
region 50-7 (25 to greater than 50 micrograms/1) . Variation
within other regions frequently approaches the full range of
phosphorus variation shown on the map (e.g. region 50-9).
Omernik attributes this pattern to the underlying geological
formations. This example shows the options open in mapping
13
-------
spatial variation, and the existence of arguments both for and
against th« regional model. In cases like this it is important
to have access to data quality statements, for example an index
expressing the percentage of information lost, by grouping (or
clustering) heterogeneous observations into uniform regions.
On the one hand, since the concept has nov been used in
some states one could argue that if further development is .
necessary, it could-be done by the states. A oareful analysis,
however, demonstrates that there are a number of remaining
research questions, many of which are more efficiently addressed
at the national level. Also, some national coordination will
help states to use a consistent rationalization process and solve
the need for states to cooperate on shared ecoregions.
Therefore, EPA is the logical source.for further development.
3.2.7 State Water Quality Uses.
The successful use of ecoregions by several states is a
strong testimonial to the usefulness of ecoregions at the state
level. Arkansas adopted an ecoregional framework to identify
natural differences in existing and achievable chemical quality
and biotic assemblages in streams.
The Ohio EPA has used the ecoregional framework to develop
biological criteria. The Minnesota Pollution Control Agency used
an ecoregional framework to summarize existing'lake chemistry and
derive appropriate achievable regional goals and criteria for
lake quality. The ecoregional framework is also a convenient and
effective way to organize, present, and interpret lake and stream
water quality information.
A variety of data analyses can be used to test/evaluate the
correspondence between ecoregions and spatial patterns of data -
dot maps, boxplots, multivariate ordination (e.g., principal
components analyses, detrended correspondence analysis) and
species profiles. These techniques have been used to evaluate
ecoregions in a qualitative way (i.e., do data distributions look
different for the regions?). See for example Lyons (1989) and
Hughes et al. (1987). Some of these techniques can also be used
to more rigorously test the ecoregions in a statistical manner
that incorporates an acceptable level of uncertainty. One has to
make sure the; scale and resolution of the ecoregions fit the
application. At the national level, less resolution means givinq
up some of the scientific precision; however, the larger
geographic ecoregions are needed for a national framework.
EPA has a significant role to help states and other agencies
to perform similar evaluations. The Ecoregion Concept is being
adopted by several states, and extended to lower levels of
generalization (smaller regions). The Subcommittee recommend*
that SPA provide advice on methods for determining regions at
14
-------
lower levels, defining reference areas and also for validating
them through statistical and other tests. EPA should provide
technical support, training and research to help states develop
local regionalizations.
3.3 Applications
The Subcommittee was asked to identify some current and
future applications for the Ecoregion Concept.
In order to illustrate the appropriate applications for
regionalizations, it would be useful to have examples of
ecological problems for which Omernik's ecoregions and other
ecoregions are not useful, to contrast with the appropriate
applications in the documentation (Gallant et al.r 1989).
Several additional pieces of information about the variables used
to delineate the boundaries would make the maps more useful. The
scale and accuracy of the individual variables should be linked
objectively to the applications. What resolution of soil mapping
is needed to support a particular number of eventual regions?
The weights given to each of the variables should be made
explicit in the documentation for the regionalization.
The Subcommittee concluded that the Ecoregion Concept has
many potentially useful applications. It could be the scientific
basis for defining the quality or characteristic of an ecosystem
within a particular area that is not influenced by anthropogenic
stress. This requires further development of the reference site
concept within a region and a better understanding and array of
biological criteria or parameters to be measured. However, the
Subcommittee feels that the concept can be used now by states in
their programs to make greater use of biological criteria as a
water quality management tool, if they have valid estimates of
natural variability, procedures for selecting reference sites,
and subdivide the ecoregions to resolve boundaries or indicate
the confidence in boundaries. The existence of the omernik
ecoregions map provides a useful initial framework for water
quality and some resource management issues. Of course*with time
and use, boundaries may be adjusted, added, or deleted, and other
regionalizations may prove more useful.
Ecoregions provide a geographic context for defining
biological criteria.' The Subcommittee supports the concept of
developing biological criteria as complements to chemical
criteria and toxicological evaluations in managing the quality a
aquatic resources. However, the Subcommittee recommends that th«
metrics and technical guidance for establishing biological
criteria (which are suggested guidelines) be carefully review**
prior to their implementation as water quality standards (which
are regulatory requirements). While diagnostic metrics like the
Index of Biotic Integrity and Index of Community Integrity (two
15
-------
metrics discussed during the review meeting)> nay prove to be
useful and valid as biological criteria, an in depth analysis of
the scientific basis and geographic limitations of these and
other candidate metrics are needed. Just as the guideline for
developing national water quality criteria benefited by in depth
scientific review, it is likely that the same will be true for
the biological criteria concept. . .
Some of the other potential areas of application include the
following:
1. Setting priorities and compare resources.needed for
monitoring different types of ecosystems,
2. Providing a framework for a "Natural History" inventory
of the Nation's ecosystems,
3. Selecting monitoring sites, i.e., in the Environmental
Monitoring and Assessment Program, ecological study sites,
bioengineered microorganism releases, new pesticide-registration,
etc, . . .
4. Estimating the potential for restoration and remediation
within geographic areas,
5. Transferring studies and ecological understanding from
one ecoregion to another ecoregion to save the cost of
duplicating studies,
6. Providing an easier means of explaining ecosystems to
the Public and a holistic method of looking at ecosystems for
environmental managers,
7. Using biological criteria in the concept of resource
management to integrate all perturbations (point source, non-
point source, habitat destruction, etc.) into a single
quantifiable parameter,
8. Providing a framework for (1) assessing national issues,
(2) setting national goals, (3) planning resource use, and (4)
summarizing and reporting accomplishments.
The SAB subcommittee encourages support by the EPA for
ecoregioa programs to better assure their development and
appropriate use. It is the SAB subcommittee's expectation, that
ecoregions will be a valuable adjunct to existing water
quality management tools, if it is properly supported with
technical guidance and user assistance.
4.0 RESEARCH NEEDS
Although the Subcommittee feels that the ecoregion eoncapt
16
-------
is an effort that will ultimately contribute to achieving water
goals in the United States, we recommend that its implementation
into the regulatory process be done gradually along with
supporting research. Our suggestions on research needs follow:
1. Develop a pilot project to determine whether defined
ecoregions perform better than less sophisticated
regionalizations for environmental management purposes.
2. Research on quantifying the delineation of ecoregions
3. Research on application of the ecoregion concept to
large rivers, lakes, and estuaries.
4. Workshops and publications of ecosystem applications to
broaden awareness.
5. Develop quantitative techniques for evaluation of
ecoregion concepts for specific applications.
6. Research in support of developing criteria for selecting
reference sites.
7. Research to develop methods to better describe
variability within an ecoregion and the effects of perturbations.
8. Research on identifying biological endpoints or criteria
and their natural variability, sensitivity, and response time.
9. Research on describing or depicting the uncertainty of
boundary locations.
The Subcommittee recommends that EPA reassemble the team
which has developed the Ecoregion Concept with a viable, well-
funded program to advance the knowledge they have already
developed and to provide technical support to users.
17
-------
5.0 SDMMAR7 OF RECOMMENDATIONS
Overall, the Subcommittee finds that Eooragiona and
regionaliiation ara valuable concepts for environmental
management and pollution control, and the Subcommittee encourages
the Agency to resume development and testing of the concept to
assess its strengths and limitations for further applications.
EPA needs to demonstrate not only that its Ecoregion Concept is
useful but that it outperforms other forms of regionalization in
its particular application areas. It is important to find out
where ecoregions perform well and where they do not, and it is
worth expending significant resources on this effort. The
Subcommittee recommends that EPA develop a pilot project to test
the performance of the defined ecoregions against other
regionalization schemes.
The Subcommittee recommends that the metrics and technical
guidance for establishing biological criteria be carefully
reviewed prior to their implementation in establishing
biologically based water quality standards.
The Subcommittee is concerned that selection criteria are
not sufficiently specific to encourage consistent and unbiased
selection of regional reference sites. The criteria should.
include the use of a large number of sites within the region to
get a reasonable estimate of natural variability.
The regionalization process should include a formal
objective protocol, with explicit QA/QC standards, and associated
quantification expected performance of a regionalization map.
Very little has been done, to date, in the direction of
accomplishing this.
The Subcommittee concluded that the Ecoregion Concept has
many potentially useful applications if proper guidance is
provided, particularly for state users. Ecoregions could be the
scientific basis for defining the highest level of quality that
is attainable by an ecosystem, if the reference site concept
within a region is refined and other recommended research is
completed. Guidance should also explain special problems, such
as how to treat large rivers that cross ecoregion boundaries.
The Subcommittee also recommends that EPA 1) initiate
additional research, an aggressive transfer of technology to
users through workshops, technical support documents, and
guidancerand 2) continue demonstrations of ecoregion
applicability. The Subcommittee concludes that the Ecoregion
Concept has applications to states for water quality management.
A specific ecoregion evaluation plan should be incorporated
18
-------
into the EMAP program, a program which will have a broad network
of monitoring stations established throughout the United States.
The EPA team which has developed the ecoregion approach must be
held together with a viable, well-funded program to advance the
knowledge it has already developed. Present ecoregion boundaries
should be subdivided by higher resolution boundaries when
necessary to produce maps at the appropriate resource level for
an assessment or research project.
While some of the above comments may appear critical, they
should not be taken as a judgment on the quality of the
ecoregions research at EPA, or on the underlying validity of the
ecoregions" concept. On the contrary, given the level of funding,
the accomplishments of the ecoregions research effort are
excellent..
19
-------
6.0 LXTBBATURB CXTBD
Bailey, R.G. (1976) Ecological Regions of the United States.
Map (scale 1:7,500,000). USDA Forest service, Intermountain
Region, Ogden, Utah. . "
Clark, S.E., D. White, and A.L. Schaedel. Submitted.. .Oregon
Ecological Regions and Subregions for Water Quality Management.
Submitted to Environ. Man.
Crowley, J.M. (1967) Biogeography. Canadian Geographer 11:
312-326.
•, • * *
Gallant, A.L., T.R. Whittier, D.P. Larsen, J.M. Omernik, and R.M.
Hughes. (1989) Regionalization as a Tool for Managing .
Environmental Resources. EPA/600/3-89/060. US EPA Environmental
Research Laboratory, Corvallis, Oregon.
Haggett, P., A.D. Cliff, and A. Frey..(1977) Locational Analysis
in Human Geography: Vol. II Locational Methods (Arnold. London),
Chapter 14 "Region Building". .
Heiskary, S.A., C.B. Wilson, and D.P. Larsen. (1987) Analysis
of Regional Patterns in Lake Water Quality: Using Ecoregions for
Lake Management in Minnesota. Lake Reserv. Man. 3: 337-344.
Heiskary, S.A. and w.w. Walker, Jr. (1988) Developing
Phosphorus Criteria for Minnesota Lakes. Lake Reserv. Man.
Heiskary, S.A. and C. B. Wilson. (1989) The Regional Nature of
Lake Water Quality Across Minnesota: An Analysis for Improving
Resource Management. J. Minn. Acad. Sci. 55(1): 71-77.
Hughes, R.M., D.P. Larsen, and J.M. Omernik. (1981) Use and
Misuse of the Terms Watershed and Stream Order. Amer. Fish. Soc.
Warmwater Streams Symposium pp. 320-326. American Fisheries
Society, Bethesda, MD.
Hughes, R.M. and J.M. Omernik. (1986) Regional Reference sites:
A Method for Assessing Stream Potentials. Environ. Man. 10(5):
629-635.
Hughes, R.M. and J.R. Gammon. (1987) Longitudinal changes in _
Fish Assemblages and Water Quality in the Willamette River,
Oregon. Trans. Amer. Fish. Soc. 116(2): 196-209.
Hughes, R.M., E.Rexstad, and C.E. Bond. (1987) The Relationship
of Aquatic Ecoregions, River Basins, and Physiographic Provinces
to the Icthiogeographic Regions of Oregon. Copeia (2): 423-432.
20
-------
Hughes, R.M. and D.P. Larsen. (1988) Ecoregions: An Approach to
Surface Water Protection. J. Water Poll. Control Fed. 60(4):
486-493.
Hughes,.R.M. (1989) Ecoregional Biological Criteria.
Proceedings of .Water Quality Standards for the 21st Century.
pp.147-151. March 1-3, 1989. US EPA Office of Water,
Washington, D.C.
Hughes, >R.M. and J.M. Omernik. (1985) Chapter 5. An
Alternative for Characterizing Stream Size.' In: Dynamics of
Lotic Ecosystems Eds. T.D. Fontaine, III and S.M. Bartell. Ann
Arbor Science Press, Michigan.
Jensen, S.E., R. Ryel, and W./s. Platts. (1989) Pilot Study-
Classification of Riverine/Riparian Habitat and Assessment of
Nonpoint Source impacts, North Fork Humboldt River, Nevada.
Intermountain Research Station, Boise, ID. 250 pp. and App.
Karr, J.R., K.D. Fausch, P.L. Angermeier, P.R. Yant, and I.J. _
Schlosser. (1986) Assessing Biological Integrity in Running
Waters: A Method and Its Rationale. 111. Nat. Hist. Survey
Spec. Publ. 5. Champaign-Urbana, 111.
Larsen, D.P., J.M. Omernik, R.M. Hughes, C.M. Rohm, T.R.
Whittier, A.J. Kinney, A.L. Gallant, and D.R. Dudley. (1986)
Correspondence Between Spatial Patterns in Fish Assemblages in
Ohio Streams and Aquatic Ecoregions. Environmental Management
10(6): 815-828.
Larsen, D.P.. D.R. Dudley, and R.M. Hughes. (1988) A Regional
Approach for Assessing Attainable Surface Water Quality: An Ohio
Case Study. J. Soil Water Conserv. 43(2): 171-176.
Lyons, T. (1989) Correspondence Between the Distribution of
Fish Assemblages in Wisconsin Streams and Omernik1s Ecoregions.
Amer. Midi. Nat. 122: 163-182.
Minn. Poll. Contr. Agency (1986) Minnesota Water Quality, Water
Years 1984-1985. 305(b) Report, pp. 19-26. Roseville, MN.
Omernik, J.M. and G.E. Griffith. (1986) Total Alkalinity of
Surface Waters: A Map of the Western Region. J. Soil Water
Conservation 41 (6): 374-378.
Omernik, J.M. (1987) Map Supplement: Ecoregions of the
Conterminous United States. Ann. Assoc. Amer. Geogr. 77(1): ua-
125. " ' '
Omernik, J.M., D.P. Larsen, C.M. Rohm, and S.E. Clarke. (1988)
Summer Total Phosphorus in Lakes: A Map of Minnesota, Wisconsin.
*''" • ' "
21
-------
and Michigan, USA. Environmental Management 12 (6): 815-825.
Platts, W.S. and M.T. Hill. Submitted. Using Ecological
Classification to Determine Stream Flow Compatibility. Submitted
to American Fisheries Society, Bethesda, Md.
«
Rohm, C.M., J.w. Giese, and C.M. Bennett. (1987). Evaluation of
an Aquatic Ecoregion Classification of Streams in Arkansas. J.
Freshw. Ecol. 4(1): 127-14.
Whittier, T.R., R.M. Hughes, and D.P. Larson. (1988)
Correspondence Between Ecoregions and Spatial Patterns in Stream
Ecosystems in Oregon. Can. J. Fish. Aquatic Sci. 45: 1264-1278.
22 l
-------
Appendix A. Examples of Testing Ecoregions
To test the effectiveness of a regionalization scheme is
not a straightforward exercise. One approach is to evaluate the
hypothesis of randomness as illustrated by the Lyons (1989) paper
on Wisconsin fish assemblages. Lyons uses a form of cluster
analysis to group samples of fish abundance data into four
groups, using 40 samples each from brooks, creeks and small
rivers in the four major ecoregions of Wisconsin. Agreements
between clusters and ecoregions range from 37.5% to 45%, and are
all found to be significant. However the null, or alternative
hypothesis being rejected in this case is that places have been
randomly and independently assigned to regions. For this to be
possible, regions would have to be highly contorted and
fragmented. The fact that regions are compact and singly bounded
ensures a high degree of non-independence. Thus the test has
merely established that a set of compact, singly bounded regions
performs better than a set of fragmented, contorted ones, not
that the ecoregions perform better than some other set of
regions. It would not be difficult to repeat the test using
random divisions of Wisconsin into four regions of approximately
equal size, and determine whether the ecoregion division
performed significantly better in its agreement with the fish
abundance data. The chances of doing this do not seem to be very
high using Lyons' data. Despite the quoted significance levels,
comparison of the number of samples showing agreement between
fish abundance cluster and ecoregion versus the number expected
by chance (from Table 3) are shown below.
Chi-Square Values fpr Fish Assemblages in Wisconsin Ecoreaions
Brooks Observed Expected
A/CHF 2 .75 ,
B/DRT 9 6.75
C/NLF 2 1.25
D/SEP 2 1.25
Creeks
A/CHF
B/DRT
C/NLF
D/SEP
Rivers
A/CHF
B/DRT
C/NLF
D/SEP
Observed
2
2
3
9
Observed
5
4
3
6
Expected
.50
1.25
2.75
6.50
Expected
1.25
2.25
2.75
3.75
CHF North Central Hardwood Forests
DRT Driftless Area
NLF Northern Lakes and Forests
SEP Southeastern Wisconsin Till Plains
23
-------
c
*ll
i
A comparison of ecoregions to river basins would also be logical
sine* basins have b««n the regions most often used for water
resource assessments.
Gallant et al. (1989) provides another example of the
difficulty of effective testing. Figure 3-1 shows total
phosphorus levels over Ohio, along with the boundaries of'the
five ecoregions present. Because of the high level of spatial
autocorrelation of total phosphorus levels, it is unfortunately
true that almost any set of regions would show some degree of
within-region homogeneity of values. Homogeneity would be quite
strong for any regionalization which preserved the basic
structure of spatial variation in Ohio, which is strongest in a
NW-SE transect. For example, division by four parallel lines
aligned NE-SW would almost certainly perform as well as the
division into ecoregions. Here again, the problem is not to
demonstrate that ecoregions perform better than no regions, but
better than other, less well informed sets of regions, and
significantly better than a random regionalization. Tests such
as ANOVA or Chisquare, which assume independence of observations
in the null hypothesis, represent an inappropriate alternative,
since independence implies infinitely contorted and fragmented
regions.
The paper on Oregon fish populations ( Hughes, et. al.,.
1987) compares the correspondence between clustered fish samples
and two sets of regions, and comes closer to an effective
comparison between ecoregions and other regionalizations. Figure
3 (from Hughes et. al., 1987) shows fish assemblages clustered
into eight classes, and mapped over the boundaries of the eight
ecoregions present in Oregon, and the 10 physiographic provinces.
From these data it is possible to count the numbers of
occurrences of each fish cluster in each region. One indicator
of goodness of fit is the number of fish samples which do not
belong to the commonest cluster found in a given region. For
example, in the Coast Range ecoregion there are 13 samples in
fish cluster 1 and 2 in fish cluster 3. If there were perfect
correspondence between ecoregions and fish clusters we would
expect all samples in this region to fall in the same cluster,
presumably cluster 1. Thus we can infer 2 misclassifications.
For the 8 ecoregions the total number of misclassifications
on the map is 22; for the 10 physiographic provinces it is 20.
Ecoregions seem to perform no better than physiographic provinces
at predicting fish populations in this example.
Given-these examples, there seems to be a pressing need to
find areas in which ecoregions do outperform other, less
sophisticated regionalizations, and to gain a better picture of
the applications for which they perform well, and those for whxcn
they are not appropriate. In fact, the Subcommittee recommend*
24
-------
that EPA develop a pilot project now to test the performance of
the EPA ecoregions against other possible ecoregions or
regionalization schemes, before embarking on major research,
development or regionalization efforts. Such a project should
also provide valuable information for other aspects of ecoregions
and ecoregion research.
25
------- |