EPA/620/R-94/009
March 1994
LANDSCAPE MONITORING AND ASSESSMENT RESEARCH PLAN
Contributions By:
Robert V. O'Neill8, K. Bruce Jonesb, Kurt H. Riitters0,
James D. Wickhamd, Ms A. Goodmanb,
March 1994
•Oak Ridge National Laboratory, Oak Ridge, TN
bU.S. EPA Environmental Monitoring Systems Laboratory, Las Vegas, NV
Tennessee Valley Authority, Norris, TN
dDesert Research Institute, Reno, NV
The Proper Citation of this Manuscript is:
Environmental Protection Agency. 1994.
Landscape Monitoring and Assessment Research Plan.
U.S. EPA 620/R-94/009.
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NV 89193-3478
Printed on Recycled Paper
-------
NOTICE
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency under Interagency Agreements DW64935962-01-0 and DW89936104-
01-0 with the U.S. Tennessee Valley Authority and .Department of Energy, respectively, and Cooperative
Agreement CR-816385-01-0 with the Desert Research Institute. It has been subjected to the Agency's
peer and administrative review, and it has been approved for publication as an EPA document.
Mention of trade names or commercial products does not constitute endorsement or recommen-
dation for use. . .
-------
Contents
Page
NOTICE ii
EXECUTIVE SUMMARY vi
ACKNOWLEDGEMENTS. 1 viii
1.0 INTRODUCTION 1
1.1 Purpose, Content, and Organization of Plan ....1
1.2 Landscape Ecology - The Theoretical Foundation
for Monitoring Landscapes , 1
2.0 SOCIETAL VALUES, ASSESSMENT QUESTIONS, MODELS,
AND INTEGRATION WITH OTHER EMAP RESOURCE GROUPS .....3
2.1 General Approach : 3
2.2 Societal Values to be Addressed by EMAP-L 3
2.2.1 Biotic Integrity and Diversity 3
2.2.2 Watershed Integrity 4
2.2.3 Landscape Sustainability and Resilience 4
2.3 Assessment Questions , 5
2.3.1 Biotic Integrity and Diversity 5
2.3.2 Watershed Integrity 5
2.3.3 Landscape Sustainability and Resilience 5
2.4 Landscape Conceptual Models 6
2.4.1 Biotic Integrity and Diversity 6
2.4.2 Watershed Integrity 8
2.4.3 Landscape Sustainability and Resilience 8
2.5 Integration with EMAP Resource Groups , 11
3.0 INDICATORS OF LANDSCAPE STATUS AND TRENDS ...15
3.1 Indicators of Biotic Integrity and Diversity 15
3.2 Indicators of Watershed Integrity 17
3.3 Indicators of Landscape Stability and Resiliency 18
4.0 GENERAL APPROACH, DESIGN ISSUES, AND ASSESSMENTS 19
4.1 Summary of Three-Step Approach .....19
4.2 Design Issues for Implementation 19
4.2.1 Definition of Landscape Scale 19
4.2.2 Definition of Landscape Units 21
4.2.3 Indicators 21
4.2.4 Data Design Issues 23
Hi
-------
CONTENTS (Continued)
Page
5.0
4.3 Assessment approach
4.3.1 Status
4.3.2 Analyzing Landscape Change
4.3.3 In-depth Assessment of Landscape Condition and
Associations with Stressors
..24
,.26
,.30
.32
IMPLEMENTATION OF EMAP-LANDSCAPES
5.1 Research and Development
5.1.1 Landscape Values/Assessment Questions/Conceptual Models.
5.1.2 Landscape Indicators ;
5.1.3 Three-Step Monitoring Approach
5.2 Implementation of Landscape Monitoring
5.3 Collaborative Efforts Proposed by EMAP-L
5.4 Project Milestones/Activities
,.35
,.35
,.35
.35
.37
.37
.38
.39
6.0 LITERATURE CITED 47
iv,
-------
Figures
Number
Page
2.1 Conceptual Model of Biodiversity Hierarchy
2.2 Conceptual Model of Hydrologic Functions
2.3 Conceptual Model of Landscape Sustainability and Resilience
2.4 Conceptual Framework Relating EMAP-L to EMAP Resource Groups.
4.1 Proposed Landscape Monitoring Approach
4.2 Scalable Landscape Units
4.3 Change in Forest Area in Pennsylvania
4.4 Example Assessment of Landscape Status and Trends
4.5 Image Substraction for Land Cover Change Detection ...
4.6 Three-dimensional Landscape Indicator Space
....7
....9
..10
..12
..20
..22
..25
.,27
..31
..33
Tables
Number
Page
2.1
3.1
4.1
4.2
4.3
5.1
5.2
5.3
Example of a Land-Use Change Matrix
Endangered Species of the Southeastern U.S.A. by Habitat
Federal Databases Potentially Useful in Identifying Landscape Changes.
Change Detection Matrix
Reported Accuracy Assessments of Change Detection Studies
Preliminary List of Research and Development Questions
to be Addressed by EMAP-Landscapes
Status and Utility of Landscape Indicators
Schedule of EMAP-L Activities and Expected Outputs
...8
.15
.24
.30
.30
.36
.38
.40
-------
EXECUTIVE SUMMARY
INTRODUCTION
Landscape ecology is the study of the structure, function, pattern, and changes in heterogeneous land
areas. Landscapes are described by the spatial arrangements of ecological resources. Landscape pat-
terns are an important determinant of the intrinsic sustainability of certain ecological processes that, in
turn, provide ecological goods and services (societal values). The Environmental Monitoring and
Assessment Program's Landscapes component (EMAP-L) will focus on those landscape patterns that
affect flows of energy, water, nutrients, and biota. The primary focus will be on the societal benefits
derived from watershed integrity, biotic integrity and diversity, and landscape stability and resilience.
The Landscape Monitoring and Assessment Research Plan is divided into five sections which describe:
(1) the theoretical basis (landscape ecology) for monitoring landscapes and how this relates to and com-
plements the objectives of EMAP; (2) the conceptual basis of the approach, including the societal values
and assessment questions to be addressed; (3) the indicators of landscape condition; (4) the methods for
monitoring and assessing the status and trends of landscape condition; and (5) the research and devel-
opment activities needed to implement the plan.
RELATIONSHIP OF LANDSCAPE MONITORING TO EMAP OBJECTIVES
The objectives outlined in the Landscape Monitoring and Assessment Research Plan parallel the
objectives of EMAP in terms of estimating status and trends in indicators of ecological condition. There is
one important distinction: the assessments made by EMAP Resource Groups do not explicitly analyze
spatial characteristics among ecological resources because these spatial characteristics are not the pri-
mary focus of their assessment questions. EMAP-L, in contrast, has as its key objective the analysis of
status and trends in indicators of landscape condition, which includes indicators of the spatial configura-
tion of ecological resources.
Many of the techniques and operational issues relating to EMAP-Landscape Characterization (EMAP-
LC) complement those of EMAP-L. EMAP-L and EMAP-LC will continue to collaborate closely to share
advances in remote sensing and geographic information system (GIS) technologies, and to achieve
economies in implementation of the respective research plans.
PROPOSED METHODS FOR ASSESSING STATUS AND TRENDS IN LANDSCAPE PATTERNS
The plan emphasizes the use of full-coverage, remote sensing and GIS methods rather than ground
sample-based methods for use in developing statistical descriptions of landscape status and trends.
Synoptic monitoring allows aggregation of landscape data in order to develop statistical descriptions about
different landscape configurations (e.g., watersheds as contrasted with landscape pattern types).
VI
-------
An important advantage of the synoptic measures is that they allow EMAP-L to integrate its measure-
ments of landscape patterns with all EMAP Resource Groups that are investigating related ecological val-
ues at a finer scale. In so doing, each EMAP Resource Group enhances its interpretive abilities; mea-
surements taken at both the Resource Group scale and the landscape scale are important to under-
standing the condition of a specific value such as biotic integrity or diversity.
EMAP-L is proposing a three-step approach to monitoring and assessment. The two first steps are
conducted to assess status and trends of landscape indicators. EMAP-L's third step is initiated when indi-
cators suggest declining landscape condition.
The first step establishes baseline landscape condition. This is done by calculating landscape indica-
tors from remote sensing imagery, and relating these to societal values. The second step involves satel-
lite-based land cover change detection to update the status of the landscape units established in Step 1.
Step 3 involves assessments of association between landscape condition and stressors. This step is
invoked in areas undergoing significant landscape change; thresholds of landscape change necessary to
trigger a Step 3 analysis have yet to be determined.
RESEARCH AND DEVELOPMENT REQUIRED TO IMPLEMENT THE PLAN
The technical and operational issues that must be resolved in order to implement the proposed land-
scape monitoring approach include: (1) selecting operational landscape units and scales for conducting
assessments; (2) refining the assessment questions that relate the spatial pattern of landscape features
to societal values; (3) refining models of landscape patterns and indicators of landscape condition; (4)
establishing accuracy objectives for assessment and data quality; (5) selecting and testing indicators of
landscape condition; (6) refining criteria to determine when indicators of status, change, and trend warrant
evaluation with stressor data (i.e., evaluation of association of observed landscape condition with envi-
ronmental stressors); and (7) developing a strategy for implementing an EMAP-L program nationwide.
EMAP-L will evaluate the proposed approach by using Landsat Multi-Spectral Scanner (MSS) data to
assess 20-year landscape status and change in selected regions. Landsat-MSS data from the early
1970s, mid-1980s and early 1990s are being compiled nationally by the North American Landscape
Characterization (NALC) Pathfinder program. Depending on results of evaluations, EMAP-L may conduct
retrospective landscape status and trends assessments with a baseline of the early 1970s.
EMAP-L proposes to address these questions through tests of the landscape monitoring approach at
pilot areas in the Mid-Atlantic Region-Chesapeake Bay area, and in the western United States (to be deter-
mined).
IMPLEMENTATION OF EMAP-L
EMAP-L proposes to implement landscape status and trends assessments within pilot areas for those
indicators, ready for implementation. Pilot areas will then be expanded to permit reporting by both natural
and standard Federal regions within the general area. As landscape indicators pass through sensitivity
and uncertainty analyses, they will be added to status and trends assessments. An overall implementa-
tion strategy for EMAP-L will be developed within the next three years.
VII
-------
ACKNOWLEDGEMENTS
We would like to thank members of the review panel, Dr. David Sharp, Dr. Sandra Turner,
Dr. Louis Iverson, Dr. John Thomas, Dr. Patrick Bourgeron, and Dr. Valerie Cullinan for their constructive
comments on the research plan.
We gratefully acknowledge the reviews of the following individuals: Dr. Bruce Milne, Dr. Chuck
Hutchinson, Dr. Frank Davis, Dr. Anthony Olsen, Mr. George Hess, Dr. Dave Mouat, Dr. Dan McKenzie,
Dr. Walter Whitford, Dr. Wayne Marchant, Dr. Ray Czaplewski, Dr. Hal Kibby, Dr. Robert Smith,
Dr. Llewellyn Williams, and Dr. Susan Franson.
The authors thank the staff of ATA for the work on the plan and Ruth Christiansen for administra-
tive support.
VIII
-------
1.0 INTRODUCTION
1.1 PURPOSE, CONTENT, AND
ORGANIZATION OF PLAN
The Landscape Monitoring and Assessment
Research Plan presents the theoretical basis for moni-
toring landscapes and describes a three-step approach
for assessing the status, changes, and trends in land-
scapes nationwide. The plan also highlights research
and development needed to implement the program
and suggests pilot studies to evaluate, test, and
demonstrate the proposed three-step monitoring
approach. There are several operational issues, such
as data acquisition schedules and data management,
that must be resolved in order to implement this
approach nationally. However, EMAP-L concluded
that the technical approach should be evaluated first,
followed by resolution of operational and implemen-
tation issues. Therefore, the primary aim of this
research plan is to highlight concepts, issues, and
approaches related to landscape monitoring and
assessments. Operational issues will be treated more
fully hi an implementation plan for EMAP-L.
The plan is organized into five sections. This sec-
tion provides background on the objectives and antic-
ipated contributions of EMAP-L. Section 2 discusses
issues related to application of EMAP's top-down
approach, specifically the relationship between soci-
etal values, assessment questions, and conceptual
models. Section 2 also provides a brief description of
how landscape monitoring might be integrated with
monitoring of other EMAP Resource Groups. Section
3 discusses landscape indicators being considered by
EMAP-L. Section 4 describes EMAP-L's three-step
strategy and proposed methodology for monitoring
landscapes. Section 5 describes research and imple-
mentation strategies for the program.
1.2 LANDSCAPE ECOLOGY - THE
THEORETICAL FOUNDATION FOR
MONITORING LANDSCAPES
In the last decade, the field of landscape ecology
has been recognized within the U.S. as a key integrat-
ing scientific discipline needed to make the concept of
ecosystem management viable and operational. For
example, many Federal agencies (e.g., U.S. Forest
Service, Bureau of Land Management, the U.S. Fish
and Wildlife Service) with land management respon-
sibilities have landscape ecologists on their staff, and
the discipline's role in agency activities is evident
within their organizational structure. Similarly, recent
public debate regarding the U.S. Forest Service's
management of forest lands in the Pacific Northwest
resulted in a landmark application of the principles of
landscape ecology and ecosystem management to
assess forest ecosystem health and to develop alterna-
tive management strategies (Everett et al. 1993).
The phrase "landscape ecology" was coined in
1939 by German geographer Carl Troll, who made
widespread use of the new technique of aerial photog-
raphy. Troll intended that his phrase, landscape ecol-
ogy, would distinguish his proposed approach for
using such imagery to interpret the interaction of
water, land surfaces, soil, vegetation, and land use
from that of conventional photographic interpretation
and cartography (Golley 1993). It has been studied
and applied hi Europe for many decades and became
generally recognized within the U.S. about 1980.
Since then, landscape ecology has rapidly evolved as
a discipline, spurred by synergistic interactions
between remote sensing and GIS techniques and
advances in ecological theory and its field applica-
tions (Golley 1993).
Landscape ecology can be defined as the study of
the structure, function, and changes in heterogeneous
-------
land areas composed of interacting organisms (Jensen
and Bourgeron 1993). Landscape ecology may be con-
sidered as the study of the interaction between land-
scape patterns and ecological processes, specifically
the influence of landscape pattern on the flows of
water, energy, nutrients, and biota (Turner 1989).
What distinguishes landscape ecology from the many
separate disciplines that it embraces (e.g., geography,
biology, ecology, hydrology) is that it provides a hier-
archical framework for interpreting ecological struc-
ture, function, change, and resiliency at multiple scales
of inquiry.
Hierarchy theory states that landscapes are orga-
nized into patterns within a hierarchy of spatial and
temporal scales. Numerous ecological and anthro-
pogenic disturbances (e.g., flooding, fires, clearing of
vegetation) maintain landscape patterns or set into
motion the creation of new landscape patterns. These
disturbance events also occur across a range of spatial
and temporal scales (Jensen and Everett 1993).
Traditional measures to protect the environment —
such as preventing water pollution or protecting biodi-
versity — have tended to focus on specific effluent
discharges or fine-scale habitat requirements. This
method has been described as the "fine-filter"
approach. In contrast, the "coarse-filter" approach to
resource conservation states that "by managing aggre-
gates (e.g., communities, ecosystems, landscapes), the
components of these aggregates will be managed as
well" (Bourgeron and Jensen 1993). In other words,
the most cost-effective strategy to maintain the
resiliency and productivity of ecological systems is to
conserve (or restore) the diversity of species, ecosys-
tem processes, and landscape patterns that create the
systems (Jensen and Everett 1993, Bourgeron and
Jensen 1993). Applying this "coarse-filter" manage-
ment method requires that landscape patterns be eval-
uated at multiple spatial and temporal scales, rather
than simply at the traditional scales of stream reach or
forest stand (Jensen and Everett 1993, Milne 1993,
MinshaU 1993, Hann et al. 1993, Bailey et al. 1993).
Hierarchy theory allows us to integrate multiple
scales of information to determine whether landscape
patterns are sufficient to allow ecological processes to
operate at the necessary scales. The objective is to
investigate changes in the distribution, dominance, and
connectivity of ecosystem components and the effect
of these changes on ecological and biological
resources (Turner and Ruscher 1988, Golley 1989,
Forman 1990, Saunders et al. 1991). Ecosystem frag-
mentation has been implicated in decline of biological
diversity and ecosystem sustainability at a number of
spatial scales (Forman 1990, Flather et al. 1992, Soule
et al. 1992, Van Der Zee et al. 1992, Wilson 1988).
Determining status and trends in the pattern of land-
scapes is critical to understanding the overall condi-
tion of ecological resources (Urban et al. 1987, Turner
1989, Forman 1990, Gosselink et al. 1990, Graham et
al. 1991, Schlosser 1991, O'Neill et al. 1992a).
Landscape patterns thus provide a set of indicators
(e.g., pattern shape, dominance, connectivity, configu-
ration) that can be used to assess ecological status and
trends at a variety of scales.
A hierarchical framework also permits two impor-
tant types of comparisons: (1) to compare conditions
within and across landscapes and (2) to compare con-
ditions across different types of ecological risks. Such
ecological risks include, for example: the risk of ero-
sion, the loss of soil productivity, loss of hydrologic
function, and expectations regarding the conservation
of biological diversity (Hann et al. 1993).
Applying the principles of landscape ecology
requires an understanding of the natural variability of
landscape patterns and processes across both space
and time. Estimates of this variability are essential to
determining whether the current condition of a land-
scape is sustainable, given its historic patterns and
processes (Jensen and Everett 1993). Moreover,
descriptions of landscape variability have proven
extremely useful in both broad-level assessment of
risk to resources, as well as to finer-scale assessments,
similar to those ongoing or planned by EMAP
Resource Groups (Jensen and Everett 1993, Hann et
al. 1993, Shlisky 1993).
Although the objectives of EMAP-L parallel the
objectives of the EMAP Resource Groups, EMAP-L
seeks specifically to:
1) Estimate, on a regional basis and with known con-
fidence, the current status, trends, and changes in
selected indicators of the Nation's landscapes.
2) Estimate with known confidence the geographic
coverage and extent of the Nation's landscapes pat-
terns and types.
3) Seek associations between selected indicators of
natural and anthropogenic stressors and indicators
of landscape condition.
4) Provide statistical summaries and periodic assess-
ments of the condition of the Nation's landscapes.
-------
2.0 SOCIETAL VALUES, ASSESSMENT QUESTIONS, MODELS, AND
INTEGRATION WITH OTHER EMAP RESOURCE GROUPS
2.1 GENERAL APPROACH
EMAP's top-down approach is used by EMAP-L
as the framework for testing and implementing a
landscape monitoring approach. This framework is
based on the following steps:
1) identify societal values related to landscape eco-
logical patterns and processes;
2) formulate assessment questions that relate to
societal values;
3) develop conceptual models that relate values to
ecological function, structure and composition,
such that candidate indicators can be identified;
4) identify candidate indicators and sampling
designs that relate to assessment questions;
5) test selected indicators by evaluating each against
a set of criteria through simulations, field tests,
and other analyses;
6) select indicators and sampling design and imple-
ment them;
7) continually reassess implemented indicators and
sampling designs and bring in (or substitute) new
, indicators (where and when appropriate).
The key to this process is linking societal values
and associated assessment questions to indicators
through conceptual models. This increases the prob-
ability that data collected on landscape condition are
relevant to societal values and key ecological compo-
nents and processes of landscapes.
2.2 SOCIETAL VALUES TO BE
ADDRESSED BY EMAP-L
The motivation for monitoring status and trends in
landscapes comes from our desire, as a society, to
maintain a wise stewardship of the environment. Our
stewardship focuses on specific "values," i.e., proper-
ties of the intact landscape that provide services to
society and that we wish to maintain. Preserving
these values, in turn, requires that we understand the
linkage between these services and the spatial pattern
of landscapes.
Ecological processes occur within the context of,
and are dependent on, the scaled spatial pattern of
landscapes (Forman and Godron 1986, O'Neill et al.
1991a,b). If the pattern is disrupted, the underlying
biotic processes that depend on the spatial pattern
will be disrupted. We develop these concepts by
examining three landscape values and showing how
preserving these values can be linked to monitoring
specific aspects of landscape pattern.
2.2.1 Biotic Integrity and Diversity
Because they provide aesthetics, recreation, and
life-support, society values intact biotic communi-
ties. Society must also preserve a diversity of genet-
ic material for future biotechnologies, medical appli-
cations, and stability.
The link between landscape pattern and intact
ecological communities is well established. In fact,
community interactions often produce the pattern.
Levin (1976, 1978) showed than predator-prey inter-
actions, combined with spatial movement of the pop-
ulations, can result in patchy spatial distributions.
Paine and Levin (1981) demonstrated that distur-
bance-recovery also produces a patterned landscape.
-------
In turn, spatial pattern impacts the way consumers
move on the landscape (Wiens and Milne 1989) and
utilize resources (O'Neill et al. 1988b). Species
coexistence also depends on spatial pattern (Shmida
and Ellner 1984). Dispersal processes interact with
pattern to separate competitors in space (Comins and
Noble 1985, Geritz et al. 1987) and permit coexis-
tence. This relationship has been shown for both ani-
mals (Kareiva 1986) and plants (Pacala 1987).
The most important cause of species loss and sub-
sequent reduction in species diversity is the loss of
habitat. As critical habitat is lost, the remaining
habitat becomes fragmented and isolated. Frag-
mentation increases with the loss of corridors
between patches of natural habitat (Forman and
Godron 1986, Harris and Gallagher 1989).
Connected habitat allows exchange of genetic
material among local populations. As corridors are
lost and habitat becomes disconnected, local popula-
tions can become extinct due to disturbance
(Saunders et al. 1991, Lovejoy et al. 1986, Wiens
1985). As a result, the simplest way to monitor
potential change in biodiversity is to measure trends
in the patch distribution of natural vegetation cover
through time.
2.2.2 Watershed Integrity
Another societal value is the landscape's ability to
collect, retain, store, and purify water. In addition,
intact landscapes control flooding and conserve soil.
Decrease in natural vegetation across a landscape,
therefore, indicates a potential for future water qual-
ity problems (Hunsaker et al. 1992).
Land cover on watersheds and hydrologic
processes are linked closely (reviews by Whitmore
and Ice 1984, Levine 1992). The physical relation-
ships have been established for more than a decade
(Donigian et al 1977, Knisel 1980, McElroy et al
1976, Anonymous 1980).
The theory has been well tested hi multivariate
empirical studies (see reviews by Berry and Sailor
1987, Regan and Fellows 1980, Levine 1992). The
empirical studies often are linked with the Universal
Soil Loss Equation (DelRegno and Atkinson 1988,
Shanholtz et al. 1988, Hession and Shanholtz 1988,
Levine and Jones 1990) to deal with erosion.
Changes hi land cover adjacent to streams have an
immediate impact on the efficiency of these buffer
strips (Phillips 1989a,b, Magette et al. 1989, Naiman
and Decamps 1990). The linkage between these
intact riparian zones and water quality is well-estab-
lished at the watershed level (Karr and Schlosser
1978, Schlosser and Karr 1981a,b, Lowrance et al.
1983, 1984, 1985, Peterjohn and Correll 1984).
2.2.3 Landscape Sustainability and
Resilience
Because of the many services rendered by intact
landscapes, society values an environment that main-
tains its own integrity (i.e., sustainable) and recovers
from disturbances (i.e., resilient). Discovery of the
relationship between landscape pattern and sustain-
ability is a relatively new development in landscape
ecology (Turner 1987b, 1989). However, percolation
theory (Gardner et al. 1989) provides an important
link between landscape connectivity and the poten-
tial for disturbance spread (White 1979, Runkle
1985). Epidemiology theory can be combined with
percolation theory to calculate the probability that a
disturbance will spread or become endemic (O'Neill
et al. 1992a).
As pattern is disrupted and landscape fragmenta-
tion increases, distances increase between disturbed
sites and source areas. The source areas provide
seeds and reservoirs of animal populations needed
for recovery. We know that northern hardwoods take
60-80 years to replace biomass and nutrients lost in
harvesting (Likens et al. 1978). This recovery time is
significantly increased if distances to seed sources
are increased or erosion sets in. Therefore, resilience
can be closely related to the frequency distribution of
distances between patches.
We know from tragic experience in the American
plains and the African Sahel that critical thresholds
exist hi landscape pattern. Beyond these thresholds,
cascading effects or positive feedbacks cause bifur-
cations that move the system into undesirable modes
of operation (Schlesinger et al. 1990).
Empirical studies (Turner et al. 1991, O'Neill et
al. 1991a,b) have demonstrated that landscapes show
a pattern at multiple scales. Disruption of this scaled
structure means that ecological processes dependent
on a particular scale have been disrupted. For exam-
ple, Rolling (1992) has established a relationship
between scales of pattern and guilds of vertebrates.
Monitoring the status and trends hi these spatial
-------
scales thus has direct implication for the sustainabili-
ty of vertebrate community structure.
2.3 ASSESSMENT QUESTIONS
Assessment questions bound the scope of the land-
scape monitoring program and define the components
of each value to be assessed. For example, biotic
diversity is a broad societal value; formulation of
assessment questions refine the components of biotic
diversity to be investigated by EMAP-L. The follow-
ing are examples of broad assessment questions that
EMAP-L proposes to address relative to each land-
scape value:
What are the proportions and geographic distribu-
tions of landscapes with acceptable biotic integrity and
diversity? How are these changing?
What are the proportions and geographic distribu-
tions of watersheds with acceptable ability to collect,
retain, store, and purify water? How are these chang-
ing?
What are the proportions and geographic distribu-
tions of watersheds (or other natural landscape units)
with acceptable landscape stability and resilience?
How are these changing?
The following are examples of a more specific set
of assessment questions that relate to each landscape
value. This set of assessment questions helps bound
the scope of landscape monitoring as questions are
more closely related to indicators of landscape condi-
tion.
2.3.1 Biotic Integrity and Diversity
What is the number, area, and distribution of land-
scape types by region?
What are the relative proportions of forest, grass-
land, etc., by landscape type and how have they
changed between t1 and tz? What is the distribution of
change?
What are the relative proportions of natural and
man-made landscapes by region and how have these
proportions changed? What is the distribution of
change?
What proportion of landscapes has significantly
changed connectivity and what is the distribution of
those landscapes?
What proportion of landscapes has significantly
changed shape complexity and what is the distribution
of those landscapes?
What proportion of landscapes has lost more than
25 percent of one cover type and what is the distribu-
tion of those landscapes?
What proportion of landscapes has altered frequen-
cy distribution of habitat patches and what is the dis-
tribution of those landscapes?
2.3.2 Watershed Integrity
What is the current status of watersheds, by region,
relative to their ability to store and convey water, con-
trol floods, and conserve soils?
What percentage of watersheds has increased/
decreased their ability to store and convey water sup-
plies from tj to t2 and what is the distribution of those
landscapes?
What percentage of watersheds has increased/
decreased their ability to control floods and conserve
soil from t1 to t2 and what is the distribution .of those
watersheds?
What percentage of watersheds has diminished
water storage capacity of those watersheds and what is
the distribution of those watersheds?
What percentage of watersheds has greater than a
critical area of impervious land cover and what is the
distribution of those watersheds?
What percentage of watersheds has potential for
erosion exceeding tolerable limits and what is the dis-
tribution of those watersheds?
2.3.3 Landscape Sustainability and
Resilience
What are the proportions of landscapes with signifi-
cantly reduced sustainability due to loss of natural land
cover and what is the distribution of those landscapes. ?
Which cover types have undergone the greatest
change in connectivity from t1 to t2 and what is their
distribution?
What proportion of landscapes is approaching criti-
cal thresholds that endanger sustainability and what is
the distribution of those landscapes?
-------
What proportion of landscapes has lost critical
scales of pattern and what is the distribution of those
landscapes?
What proportion of landscapes has altered recov-
ery ability by critically increasing distance to seed
sources and what is the distribution of those land-
scapes?
Finally, EMAP-L will develop a specific set of
assessment questions relative to associations between
landscape condition and stressors, as well as to asso-
ciations between landscape condition and other eco-
logical resources (e.g., plant and animal richness).
The following are examples of these types of ques-
tions:
What is the spatial distribution of changes in cover
type and are changes associated with changes in plant
and animal species richness?
What is the relationship between landscapes with
low connectivity of natural cover types and forest pro-
ductivity?
What is the relationship between landscapes with
low connectivity of natural cover types and stream
benthic condition?
Wliat is the relationship between landscapes with
low connectivity and the type, distribution, and mag-
nitude of land use ?
What proportion of landscapes has significantly
increased probability of pest dispersal and what is the
distribution of those landscapes?
What proportion of landscapes has significantly
increased risk of disturbance (e.g., fire) and what is
the distribution of those landscapes?
In order for EMAP-L to address conditions of
landscapes relative to each value, EMAP-L must
establish a set of thresholds relative to landscape
indicators. For example, thresholds for indicators of
watershed integrity must be established in order to
identify those watersheds that have integrity versus
those that do not. EMAP-L recognizes that these
thresholds may vary between regions of the United
States. Identification of condition thresholds for
landscape indicators is a primary research area for
EMAP-L.
2.4 LANDSCAPE CONCEPTUAL
MODELS
Conceptual models link indicators to societal val-
ues. The definitions or descriptions of the societal
values (biotic integrity and diversity, watershed
integrity, and landscape sustainability and resiliency)
provide some insight into measurements that could
be used to represent them. For example, defining
watershed integrity as the ability to collect, retain,
store and purify water suggests that measurements of
soil and vegetation will be used to support watershed
integrity. The conceptual models provide a frame-
work for identifying the specific indicators to be
measured and scales of concern. A general concep-
tual model is illustrated for each landscape value list-
ed below. We propose to refine these models through
a series of workshops, as well as from data gathered
in pilots (see Section 5).
2.4.1 Biotic Integrity arid Diversity
. Biodiversity is defined by the Office of
Technology Assessment (1987) as "the variety and
variability among living organisms and the ecologi-
cal complexes in which they occur." This is a com-
monly cited definition of biodiversity (Noss 1990,
see also Stoms and Estes 1993). It is insightful
because it acknowledges the value of ecological
complexes in their own right not necessarily linked to
living organisms. Examining the diversity of ecolog-
ical complexes is the appropriate level of organiza-
tion for EMAP-L.
Our conceptual model of biodiversity (Figure 2.1)
is from Noss (1990). It presents biodiversity as four
levels of organization (genes, species, communities,
and landscapes) divided into compositional, structur-
al, and functional components, which are the prima-
ry components of an ecosystem (Franklin et al.
1981).
The conceptual model helps to identify the specif-
ic indicators to measure landscape biodiversity.
Landscape composition and configuration influence
the number and types of plants and animals that can
inhabit an area (Saunders et al., 1991). Indicators of
landscape composition include number and propor-
tion of cover types, including the number and pro-
portion of natural cover types in a watershed. The
presence of riparian habitat is likely "keystone" in
-------
genetic
processes
demographic processes,
life histories
interspecific interactions,
ecosystem processes
landscape processes and
disturbances, land-use
trends
Indicators
Composition
Structure
Function
Inventory and
monitoring tools
Regional
Indentity, distribution,
.—»•; inaemny, aisiri
Landscape richness, and
proportions of patch
(habitat) types and
multipatch landscape
types; collective
patterns of species
distributions
(richness, endemism)
Heterogeneity;
connectivity; spatial
linkage; patchiness;
porosity; contrast; grain
size; fragmentation;
configuration;
juxtaposition; patch size
frequency distribution;
perimeter-area ratio;
pattern 9f habitat layer
distribution
Disturbance processes
(areal extent, frequency
or return interval, rotation
period, predictability,
intensity, severity,
seasonably); nutrient
cycling rates; energy flow
rates; path persistence
and turnover rates; rates
of erosion and
geomorphic and
hydrologic processes;
human land-use trends
Aerial photographs
(satellite and conventional
aircraft) and other remote
sensing data; Geographic
Information System (GIS)
technology; time series
analysis; spatial statistics;
mathematical indices (of
pattern, heterogeneity,
connectivity, layering,
diversity, edge,
morphology, auto-
correlation, fractal
dimension)
Figure 2.1 Conceptual model of biodiversity hierarchy
-------
affording high biodiversity of flora and fauna.
Riparian corridors are structurally complex ecosys-
tems that support high biodiversity (Odum 1978,
Gregory et al. 1990, Naiman et al. 1993), especially
in the arid, western United States (Thomas et al.
1979). Indicators of structure include heterogeneity
(e.g., the number of different natural cover types sur-
rounding a point), edge-to-area ratios, and inter-
patch distance (by cover type).
Measures of landscape function are admittedly
less straightforward, primarily because they involve
measurement over time or some predictive ability.
Measures of percolation, e.g., the measure of con-
nectivity of a cover type, (Gardner et al. 1992) permit
hypothesis testing of variability in disturbance
regimes. Measures of patch persistence will also be
useful. Loucks (1970) pointed out that the niche of a
species has a temporal domain. Land cover transi-
tion matrices can be created to determine patch per-
sistence. An example land cover transition matrix is
provided by Hall et al. (1991a) (Table 2.1).
2.4.2 Watershed Integrity
EMAP-L has defined watershed integrity as the
ability to collect, retain, store, and purify water. The
conceptual model (Figure 2.2, from B. Richter, The
Nature Conservancy, personal communications)
shows the important categories of indicators to mea-
sure on watersheds. As indicated by Figure 2.2, land
cover is the primary driving force behind a watershed
or other area of land to maintain its ability to process
water.
Slope, vegetation, and land cover provide data to
measure the ability to collect, retain, and store water.
Table 2.1 Example of a Land-Use Change Matrix
Aspects of collection, retention, and storage of water
can be measured by combinations of slope and land
use (e.g., agriculture on excessive slopes).
Agriculture on slopes greater than three percent
increases risk of erosion (USDA 1951). Erosion esti-
mates are available through the U.S. Soil
Conservation Service National Resources Inventory
(NRT) program.
Measurements of riparian habitat may be the most
useful gauge of natural water purification. Riparian
habitat has been shown to function as a "sponge," and
prevents excess runoff of nutrients and chemicals
into streams (Peterjohn and Correll 1984, Lowrance
et al. 1984, Schlosser and Karr 1981a,b). Peterjohn
and Correll (1984) note that the ability of riparian
forests to retard and store excess runoff is likely a
universal effect. Simple presence of vegetation adja-
cent to water is a potential indicator of all four
aspects of watershed integrity (collection, retention,
storage, and purification).
2.4.3 Landscape Sustainability and
Resilience
The conceptual model for landscape sustainability
and resilience is based on combining the three pri-
mary ecosystem attributes of Franklin et al. (1981)
with the concepts of state and transition and
resilience from Moiling (1973). The model is illus-
trated in Figure 2.3. The composition, structure, and
function add up to define the ecosystem state. Under
natural conditions, the ecosystem can be conceived
as occupying N possible states, within which it varies
over time (these are considered normal or typical
states). These N possible states comprise a dynamic
1983 State
1973 state
CIrng Rgnrt
Brdlf
Mixed
Cnfr
Other
Clearings
Regenerating
Broadleaf
Mixed
Conifer
Other
17.09
4.55
1.12
0.52
1.04
0.53
45.54
30.83
19.72
6.81
4.37
3.14
16.72
16.93
47.06
11.28
1.81
3.19
15.20
37.27
27.61
58.11
31.02
8.60
5.22
10.03
4.16
22.55
57.80
13.38
0.12
.0.36
0.28
0.72
3.93
77.06
Note: Tne italicized numbers are retention frequencies.
-------
• Physical Patterns
Physical Processes
Scale
Regional Aquifer Capacities
and Interaction with
Surface Water
Valley Fill and Floodplain
Aquifer Capacity and
Confinement
Sediment
Loading
-
Biochemical
Loading
Floodplain Deposition
and Erosion
Floodplain Disturbances
CD
I
CD
CD
D)
CD
CO
CD
•j^
CO
Source: Modified from B. Richter, Western Regional Office, The Nature Conservancy.
Figure 2.2 Conceptual model of hydrologic functions
9
-------
CO
111
CO
LLJ
O
CO
CO
HI
u.
8
CO
II _
52 co
'8,
o
I
in
O 0)
o±
CD
£ i £14 1 Si!
CD
s.
.52 o
-
Q)
. _ „
JE-o 03 Q Q..
O .0
_ »- £2
(0*= O **
^ w« S
® O-=
-------
stability (Sprugel and Bormann 1981), which, in
effect, defines sustainability. Disturbances pass
through the ecosystem over time causing the transi-
tion to a new ecological state. Resilience is the abil-
ity of the ecosystem to take on a new state within the
range of normal or typical states following distur-
bance.
The dashed line to an altered ecological state
highlights the possibility that disturbances (presum-
ably anthropogenic ones) have caused a transition to
an "altered" ecological state, and that transition back
to one of the normal or typical ecological states is
perhaps either not probable or possible.
Several indicator categories are listed under the
compositional, structural, and functional ecosystem
components, which measure aspects of landscape
sustainability and resilience as described herein. The
number and distribution of natural cover types with-
in the landscape measures an aspect of landscape
stability and resilience because imposition of barriers
between these patches hinders dispersal between the
patches. This measure of composition could be com-
bined with measures of structure such as inter-patch
distance edge to area ratio, and patch size distribu-
tion. The land cover transition matrix introduced
above under the discussion of biodiversity would
serve as an indicator of landscape sustainability and
resilience for the functional component of the
ecosystem. Indicators derived from percolation theo-
ry (Gardner et al. 1989) would also fall under the
functional ecosystem component.
Several examples illustrate how these indicators
could be used to address landscape stability and
resilience. Prior to settlement, the eastern United
States was largely forested (Whittaker 1975).
Fragmentation of forests into smaller, more isolated
patches creates a source/sink gradient for dispersal of
propagules needed to maintain succession. An abun-
dance of small forest patches at high inter-patch dis-
tances would indicate that forest successional pat-
terns could not be maintained.
Grover and Musick (1990) have shown how graz-
ing practices and climate have interacted to create
excessive soil erosion. As a result of the grazing
intensity and climate scenario, shrub encroached on
natural grasslands hi the Southwest. Shrubland
encroachment, hi turn, caused accelerated eolian ero-
sion, which has prevented the return of grasslands
(and artificially maintained shrublands) even hi the
absence of grazing pressure.
The study by Grover and Musick highlights the
introduction of an "altered" state shown hi the con-
ceptual model. Grazing pressure, soil loss, and veg-
etation were the inputs into indicators that were used
for this study. Each of these can be measured by
EMAP-L. Grazing pressure can be estimated from
the Department of Commerce, Census of
Agriculture, and livestock inventories (by county);
soil loss can be estimated from NRI data; and vege-
tation can be measured from land cover and remote
sensor data.
2.5 INTEGRATION WITH EMAP
RESOURCE GROUPS
Biotic integrity and diversity, watershed integrity,
and landscape sustainability and resiliency are
umbrella values that are relevant to both a landscape
monitoring program and the individual EMAP
Resource Groups. For example, EMAP-Forests is
interested in status and trends of indicators of forest
habitat structure relating to breeding birds. EMAP-
Surface Waters is interested hi the biotic condition of
streams. EMAP-Landscapes is interested hi land-
scape condition (e.g., pattern indicators) relating to
biotic integrity. Individual resource group values of
biotic integrity are nested within a landscape.
Indeed, streams, wetlands, forests, agroecosystems,
and arid ecosystems are all nested within a landscape
(although not all of them always in one place).
Because of this spatial nesting, landscape condition
indicators are often viewed as stressor indicators at
the individual ecological resource level. For exam-
ple, EMAP-L considers the degree of connectivity as
a landscape condition relevant to biotic diversity,
whereas EMAP-Forests might view this as a stressor
(e.g., fragmentation) associated with condition of
forests at the plot level. However, EMAP-Forests
also is interested in landscape conditions within and
around forest patches to gain greater insight into for-
est sustainability.
Despite some of these differences, status and
trends hi indicators at both scales (e.g, resource
groups and landscapes) are necessary to address biot-
ic diversity at a regional scale. Figure 2.4 provides an
example of this relationship. In this example, the
basic value is biotic diversity, but the assessment
questions specifically relate to a component of biotic
diversity, breeding bird habitat. One assessment
question deals with horizontal and vertical structure
11
-------
i
Q. £.
OT "CD
S >
OOEO.CD
Sf°
fe
^o S
. O "K r=
—' Cu ^
S-£c£
CO
co'
EC .£10. to
•2 2
3
C/l
CO
a. co
CD O O
DCQ_O
— .93
> o §
CO CD
CO
O .C O
O CO Q
I I I
CO
O -t=^-
CO i-n'42
I is
So "c > O" ™
t^oSS <
cSw
DC
o
53 -
O CD
15.=:
CM
I
O)
12
-------
at the "stand" or "patch" level and the other with spa-
tial configuration of ecological resources at a land-
scape level. The conceptual model suggests
(although not illustrated in this case) that both scales
are important hi determining the quality of breeding
bird habitat. Therefore, habitat measurements are
necessary at both scales to determine status and
trends in breeding bird habitats. In this example, a
series of common habitat measurements (or at least
those that could be normalized) could be implement-
ed among EMAP Resource Groups in order to
address the stand-level assessment questions, where-
as certain landscape metrics must be calculated in
order to address the landscape-level assessment ques-
tions. Figure 2.4 thus expands the general top-down
approach of EMAP (see earlier discussion) into a
framework for integrating monitoring approaches
among EMAP Resource Groups and EMAP-L. Such
an approach could be used to address other common
societal values.
13
-------
-------
3.0 INDICATORS OF LANDSCAPE STATUS AND TRENDS
Some indicators will be used to quantify land-
scape status and trends or identify areas undergoing
significant change. Others will be used for in-depth
assessments.
Candidate indicators represent a spectrum from
fully field-tested to preliminary concepts. We have
organized the presentation into three sections. The
criteria by which the indicators will be evaluated is
discussed hi Section 5.
3.1 INDICATORS OF BIOTIC INTEGRITY
AND DIVERSITY
Landscape monitoring can contribute to our
understanding of status and trends in biodiversity by
addressing changes in the configuration of habitats.
The simplest indicator considers the number of pix-
els that show changes hi land cover. Changes in nat-
ural vegetation cover reflect loss or gain of wildlife
habitat. Attention will be focused on specific transi-
tions, such as the loss of rare land cover types (e.g.,
sky island forests of the Southwest).
O'Neill et al. (1992b) present a list of rare and
endangered species hi southeastern United States that
are uniquely dependent on a habitat type. A portion
of the list is reproduced hi Table 3.1. In specific
regions, such as central Florida, the loss of sandy
scrub habitat permits an immediate indicator of habi-
tat status and trends relative to endangered species.
We would also be concerned with the status and
trends in indicators of landscape pattern. We will
focus on aspects of pattern that are known to influ-
ence biodiversity. For example, we will follow
changes hi landscape connectivity as reflected hi loss
of corridors between patches of natural habitat
(Forman and Godron 1986, Harris and Gallagher
1989). A number of field studies have shown that
Table 3.1 Endangered Species of the Southeastern
U.S.A. by Habitat
Sandy Pine/Oak Scrub (Central Florida)
Florida Scrub Jay (Aphelocoma c. coerulescens)
Eastern Indigo Snake (Drymarchon corals couperi)
Blue-tailed Mole Skink (Eumeces egregius lividus)
Sand Skink (Neoseps reynoldsi)
Wide-leaf Warea (Warea amplexifolia)
Four-petalled Pawpaw (Asimina tetramera)
Florida Bonamia (Bonamia grandiflora)
Pygmy Fringe Tree (Chionanthus pygmaeus)
Florida Golden Aster (Chrysopsis floridana)
Scrub Lupine (Lupinus aridorum)
Scrub Plum (Prunus geniculata)
Scrub Mint (Dicenandra fruteaceus)
Snakeroot (Erygium cuneifolum)
Lakelas Mint (Dicerandra immaculata)
Highland Scrub Hypernicum (Hypemicum cumulicola)
Papery Whitlow-wort (Paronychia charactacea)
Wireweed (Polygonella basiramia)
Carter's Mustard (Warea carter!)
Mature Oak-Hickory Hardwoods
Large-flowered skullcap (Scutellaria montana)
Prairie/Savannah Arkansas
Geocarpon minimum
Salt Marsh
Atlantic Salt Marsh Snake (Nerodia f. taeniata)
Pine-Grass Ecotone
Gopher Tortoise (Gopherus polyphemus)
Rough-leaved Loosestrife (Lysimachia asperulaefolia)
Conifer-Hardwood Ecotone
Carolina Northern Flying Squirrel (Glaucomys s. coloratus)
Scrub-Agriculture Ecotone
Florida Grasshopper Sparrow (Ammodramus s. floridanus)
Riparian-Hardwood Ecotone
Indiana Bat (Myotis sodalis)
Granite Outcrops
Granite Snapdragon (Amphianthus pusillus)
Quillwort (Isoetes melanospora)
Quiilwort (Isoetes tegetiformans)
15
-------
wildlife utilize these corridors to move between
resource patches (e.g., Mwalyosi 1991).
The length of edges adjacent to natural patches
can also be related to status of the landscape relative
to wildlife (Correll 1991, Gardner et al. 1991, Turner
et al. 1991). In general, loss of edge results in
decreased biodiversity (Wilson 1988). Edges often
form unique habitats that may be associated with rare
and endangered species (O'Neill et al. 1992b).
It is also relevant to examine the relationship of
edges to patch size. For example, cowbirds on forest
edges are nest predators on warblers (Harris 1988,
Terborgh 1992). Forest patches have to be suffi-
ciently large so that nest sites are far away from
edges and cowbirds cannot find them. If patches get
too small, warbler populations start to decline.
Therefore, it will be important to quantify status and
trends hi patch size distributions.
Much of what we understand about the influence
of landscape pattern on ecological processes is based
on patch configuration (Kareiva 1986, Franklin and
Forman 1987). For example, the frequency distribu-
tion of patch sizes can be important because some
species require a minimal patch size. As the distrib-
ution of patch sizes changes, the landscape becomes
more hospitable to some species and less hospitable
to others.
The largest habitat patch may also serve as a
reservoir that maintains a population on the land-
scape. Fragmentation of a landscape into many iso-
lated patches has been shown to reduce biodiversity
(Bierregaard 1990, Gardner et al. hi press, Saunders
et al. 1991, Lovejoy et al. 1986, Wiens 1985).
Therefore, we will follow changes in the largest
patch size.
It would also be useful to monitor the frequency
distribution of distances between patches. These
nearest-neighbor distances are related to the difficul-
ty experienced by wildlife moving across the land-
scape.
Some spatial arrangements of patches may be
particularly vulnerable to fragmentation. Isolated
habitat may be configured in a longitudinal pattern,
like a string of pearls. Examples include alpine tun-
dra along ridgetops of the Rockies, dune vegetation
along beaches, and granite outcrops. Disruption of
linear patch configurations, by removal of a single
patch, may split in two the entire habitat if the gap
exceeds the dispersal ability of the populations. We
will develop an indicator that will quantify the status
and trends of these spatial configurations.
One method to assess a land cover would be to
ask: "What is the current status of land cover, com-
pared to its potential?" This suggests an index that
compares each pixel with overlying maps of poten-
tial vegetation cover (e.g., Kiichler 1964) and calcu-
lates trends in the difference.
It is also possible to quantify status and trends in
landscape potential for specific wildlife (see
Danielson 1992). Consider a square "window," the
size of an organism's home range. Within the win-
dow we could consider a variety of habitat require-
ments, such as vegetation mixture, edge, and avail-
able water. We could then place the window over a
corner of the landscape map and determine if the land
covers within the window met all habitat require-
ments. The window could then be moved systemati-
cally over the map to obtain an overall indicator of
the status of the landscape for this organism. We
could design a suite of windows for insects, birds,
mammals, etc. This approach provides a simple indi-
cator of status and trends that could be interpreted in
terms of the impact on animals of a change in land-
scape pattern.
Another indicator of the impact of land cover
changes on wildlife would be miles of new roads. In
addition to fragmenting the landscape, roads have an
immediate impact on wildlife mortality.
Percolation theory (Stauffer 1985) provides a
framework for relating landscape pattern to the abil-
ity of an organism to move across the landscape
(Gardner et al. 1987). Diffusion rates can be calcu-
lated and interpreted in terms of wildlife utilization
or disturbance spread. The percolation backbone
defines the fewest steps needed to traverse the land-
scape.
Percolation theory also defines percolation
thresholds of habitat coverage (Gardner et al. 1992).
On a random square lattice, the critical value is 59.28
percent. If percent cover for habitat is less than this
value, the landscape becomes dissected into isolated
patches. Resource utilization scale measures the
scale at which an organism must operate to utilize the
resources on a landscape (O'Neill et al. 1988b).
Percolation theory thus suggests several indicators of
landscape status and trends.
Empirical studies (O'Neill et al. 1991a,b) have
confirmed the prediction from Hierarchy Theory
16
-------
(O'Neill et al. 1986, O'Neill 1988, 1989, O'Neill et
si. 1989) that landscapes should show pattern at dis-
tinct scales. This approach uses statistical analysis of
transect data to identify multiple scales of pattern (S.
Turner et al. 1991). Disruption of this scaled struc-
ture means that ecological processes determining a
particular scale have been disrupted. We will moni-
tor the status and trends of landscapes by quantifying
the number of scales extracted from the remote
imagery.
Recently, Holling (1992) has established a rela-
tionship between landscape scales and vertebrates.
Because of the close relationship between vertebrate
body size and home range, Holling was able to estab-
lish that clusters of body sizes can be directly related
to landscape scales. Rolling's work makes it possi-
ble to relate the loss of a landscape scale to the risk
of losing a guild of vertebrates that are dependent on
that specific scale of resource distribution.
One of the most recent, and most sophisticated,
indicators of wildlife use utilizes the concept of cel-
lular automata. This involves an imaginary organism,
or automaton, that moves across the landscape, one
pixel at a time. The automaton moves randomly, but
never moves backward. The organism steps freely
(probability = 1.0) onto natural vegetation, and
moves less freely (probability « 1.0) across clear-
ing, agriculture, or other land uses. By releasing
1,000 automata, allowing each to take 1,000 steps
and recording the number of times a pixel is visited,
it is possible to evaluate how organisms will utilize a
landscape configuration. This approach is particular-
ly valuable for identifying gaps or clearings that are
likely to be heavily used.
Another indicator of landscape status and trends
can be developed by weighting individual pixel tran-
sitions. One might, for example, apply a greater
weight to a transition that fragments a large patch.
Similarly, a transition could be weighted by the prob-
ability of forming a barrier to animal movement or
breaking up a corridor. It would be important to dis-
tinguish between 100 pixels scattered randomly and
100 pixels in a line, forming a new barrier to animal
movement.
Individual transitions can also be weighted by
characteristics of the entire landscape. Loss of a rare
cover type may be more important than loss of an
abundant cover type.
3.2 INDICATORS OF WATERSHED
INTEGRITY
It is also possible to develop indicators of status
and trends in landscapes that relate to changes in
water quality due to changes in terrestrial landscapes
(Hunsaker et al. in press). Across a region, increases
in agriculture and urban land use or decreases in nat-
ural vegetation indicate a potential for future water
quality problems. The basic cover changes could be
weighted by distance to water, soil type, and slope
calculated from digital elevation models. Essentially,
the same data set can be used with the Revised
Universal Soil Loss Equation to follow status and
trends in erosion potential.
A second type of indicator might focus on the
potential for undesirable hydrologic events. For
example, a flood indicator could include vegetation
cover and surficial geology.
Riparian zones (e.g., vegetation adjacent to
water) are important buffers for maintaining the
water quality of streams (Naiman and Decamps
1990). Changes in width of buffers, weighted by
slope, would be an important indicator. The actual
index might be average width, or miles of riparian
zone that are narrower than desirable. This standard
could be applied by counting pixels that encroach
into the recommended buffers.
Similar indicators might be the formation of con-
tiguous agriculture adjacent to a stream or lake, both
of which may contribute to reduced water quality.
We would weight more heavily a pixel change that
increases contiguous agricultural cover. Contiguous
agricultural fields along flow paths increase hydro-
logic length and lead to channelization. Specifically,
we would calculate hydrologic distance which
includes vegetation characteristics along the over-
land flow path calculated from digital elevation data.
Hydrologic pathways are altered by road sur-
faces, and water quality can be affected when the
roads intersect streams. Therefore, miles of new
roads weighted by distance to water might form
another useful indicator.
Some research indicates that more precise predic-
tions of water quality can be made on the basis of
land cover on the watershed. The approach is empir-
ical, relying on correlations between land use and
water quality on monitored streams. The correlations
17
-------
will undoubtedly be different in different regions of
the country, and considerable additional research is
needed. Nevertheless, such a watershed indicator
would have enormous potential for monitoring status
and trends in landscapes at the watershed level.
3.3 INDICATORS OF LANDSCAPE
STABILITY AND RESILIENCE
As previously discussed, percolation theory
(Gardner et al. 1989) provides a framework for relat-
ing landscape pattern to wildlife habitat. Because it
deals with landscape connectedness, the same theory
is useful for monitoring the potential for disturbances
to spread across the landscape (White 1979, Runkle
1985). Specifically, if percent cover for disturbance-
prone land cover is higher than a threshold value, the
potential for disturbance spread becomes significant-
ly higher (Gardner et al. 1991, 1992). Particularly
noteworthy is the combination of epidemiology the-
ory with percolation theory to calculate the probabil-
ity that a disturbance or pest will spread or become
endemic (O'Neill et al., 1992).
In addition to predicting economic activity, miles
of roads also indicate potentially accelerated disper-
sal pathways for pests and exotic biota. This has
been particularly noteworthy in the spread of the
gypsy moth. New infestations often begin around
trailer parks where recreational vehicles have trans-
ported the larvae from one forested area to a distant
point in the region.
Unlike wildlife habitat and water quality, the rela-
tionship between landscape pattern and disturbance
spread is a relatively new development in landscape
ecology (Turner 1987b, 1989). It can be expected,
therefore, that a number of new indicators will be
developed as research continues.
For example, we have little insight into how spa-
tial pattern affects the ability of systems to recover
from disturbance. We know that northern hardwoods
may take 60-80 years to replace biomass and nutri-
ents lost in harvesting (Likens et al. 1978). How
would this recovery time change if distances to seed
sources were increased or if erosion set in?
We need to identify ecological systems that are
particularly sensitive to spatial disturbances. We are
aware of the proverbial erosion effects of tire tracks
in the Arctic Tundra, but arid lands may be equally
sensitive. Even the casual observer can see how
small alterations in natural landform result in major
changes in arid land vegetation.
The potential sensitivity of arid lands also alerts
us to the need to identify critical thresholds in land-
scape pattern. We know from percolation theory that
small changes in land cover can critically alter land-
scape connectivity.
18
-------
4.0 GENERAL APPROACH, DESIGN ISSUES, AND ASSESSMENTS
This section will describe the general approach to
monitoring and assessing landscape condition over
time. Some implementation issues and researchable
questions will then be described, followed by an
example of the general approach.
4.1 SUMMARY OF THE THREE-STEP
APPROACH
EMAP-L will use a phased, three-step approach
of monitoring and assessment (Figure 4.1). A "base-
line" landscape condition is first established in Step
1, then Step 2 is designed to quantify change in land-
scapes over time. Satellite images are used for both
steps. Together, the first two steps are the basis for
statistical reports of landscape status and trends.
Step 2 also serves to identify those landscapes which
have changed the most and are therefore of particular
interest for assessments. Depending upon the results
of Steps 1 and 2, Step 3 entails in-depth analyses of
particular situations of interest.
Status and trends reports will be prepared, at fixed
time intervals, for all landscapes with respect to
broadly-defined societal values as measured by con-
dition indicators. The primary objective is to sum-
marize the overall condition of landscape units on a
regional basis with known confidence. To help
decide whether a more in-depth analysis is warrant-
ed, the amount and types of landscape pattern change
will be reported. EMAP-L will take advantage of
GIS technology in analyzing and displaying land-
scape status and change.
In-depth assessments will address the relation-
ships among specific subsets of societal values and
environmental stresses. They will be prepared only
when warranted by evidence of changing landscape
conditions for specific geographic areas, and the
objectives will depend upon the particular types of
change which have been observed as well as the soci-
etal values and environmental concerns that are per-
ceived to be important in those areas. By improving
the understanding of the relationships between stres-
sors and landscape condition, these analyses should
enable projections of relative risks to societal values
given different landscape condition and stressor sce-
narios.
4.2 DESIGN ISSUES FOR IMPLEMENTATION
This section will mention some design issues that
are relevant to implementing the three-step approach.
These include the definition of landscape scales and
units, the uses of different types of indicators, and the
choices of databases.
4.2.1 Definition of Landscape Scales
Scales can be defined for the spatial, temporal,
and attribute dimensions of a landscape (more gener-
ally, for any ecosystem). Selecting a spatial scale for
monitoring means defining a gram size (i.e., the min-
imum resolvable size of a unit) and extent (i.e., the
number of units) for a particular calculation or
assessment. The temporal scale is the frequency at
which landscapes will be analyzed. The attribute
scale is a function of the types and numbers of dif-
ferent attributes which are recognized in an analysis.
It is a design issue to optimize the scale for each
value, assessment question, and indicator. These
choices must consider the available data and the sen-
sitivity of indicators at different scales. Furthermore,
the same indicator may be analyzed at several differ-
ent scales (for example, over different total extent)
depending upon the assessment question, since some
19
-------
C
T3
CD
_
CO 0> O CD
-O~ O D)
,_ T3 CO-ili O
J-
S&«?
o
0)
g--
--
811
CD ^- f^
O QD J5 o
Q.
Q.
CO
O>
O
I
O
(0
•o
,55
"S
(0
2
Q.
O)
il
20
-------
ecological processes operate at multiple scales
(Saunders et al. 1991, Ledig 1992, Terborgh 1992,
Ulfstrand 1992). This rationale assumes that if dif-
ferent ecological processes operate at distinct scales,
then the scale must be matched to that process
(Meentemeyer 1989). Turning this rationale around,
it is also possible to detect the scales (and changes in
scales) at which ecological processes are operational
(S. Turner et al. 1991, O'Neill et al. 1991b).
4.2.2 Definition of Landscape Units
As for ecosystems in general, there is not a
unique definition of what constitutes a landscape.
But all numerical procedures require that a specific
area be defined, and all assessments must be made
for specific geographic areas. Thus, landscape units
will be operationally defined as appropriate for spe-
cific assessment questions and for the conceptual
models and indicators which are used to answer
those questions. It is a design issue to optimize the
landscape units for monitoring.
The measurement unit is the smallest resolvable
area on the landscape image; it is completely speci-
fied by the data collection procedures (Figure 4.2).
An example is a pixel in a raster landscape map.
Flexibility is necessary when aggregating measure-
ment units into landscape units because the aggrega-
tion strategy for one analysis will probably not be
optimal for any other analysis, even though the total
reporting region may be the same. To provide for
both multi-scale analyses and consistent reporting, it
is useful to distinguish "fixed" landscape units from
"scalable" units.
There are two types of fixed units: "reporting
units" and "natural regions." A reporting unit is a
geographic area for which statistical reports will be
prepared, for example, a standard Federal region. A
natural region is made up of one or more geographic
areas classified on the basis of similar large-scale and
long-term biogeophysical attributes, e.g. ecoregions
(Omernik 1986, Bailey 1991). The variety of poten-
tially useful classification rules means that a variety
of natural regions could be used depending upon the
assessment question.
Scalable units are needed to address landscape
ecology issues at multiple scales within a hierarchi-
cal framework. Examples of scalable units include
patches, patterns, and landscapes. A patch unit is a
set of contiguous measurement units (e.g., pixels)
which have the same numerical value. A pattern unit
is a collection of measurement units and/or patch
units which have the property of being the minimum
unit descriptor of a larger spatial area. A landscape
unit may be a collection of pattern units in the spirit
of Forman and Godron's (1986) definition of a land-
scape. More generally, a landscape unit is simply a
collection of measurement, patch, and/or pattern
units which comprise a logical grouping. Patches,
patterns, and landscapes are considered to be scalable
because their precise definitions depend upon the
choices of scales and indicators as described above.
The scales of assessment questions and indicators
suggest two types of landscape units, watersheds and
landscape pattern types (LPTs) (Wickham and
Norton 1994). Both watersheds and LPTs appear to
capture (or bound) four important flow processes
operating within and among landscapes: flows of
energy, water, nutrients, and biota. These processes,
in turn, are the main factors influencing the land-
scape values of biotic integrity and diversity and
landscape stability and resilience. Watersheds are an
obvious choice for addressing landscape condition in
terms of watershed integrity. Furthermore, both
watersheds and LPTs are scalable and hence suitable
as landscape units. Several scales of watershed and
LPTs will range from approximately 103 to 106 units
in extent, and from approximately 1 to 100 hectares
in grain size.
4.2.3 Indicators
A number of indicators were introduced in
Section 3. It was mentioned that some would be used
for assessing status and trends and others for in-depth
analyses. The exact roles that each indicator will
play remain to be worked out for most indicators. Of
particular interest is the set of indicators that will be
used in Steps 1 and 2 of the three-step process.
These are important because they are the basis for
statistical reports of status and trends. The issue is
complicated by the planned use of satellite images
for both steps because not all indicators can be esti-
mated strictly, from satellite images.
Landscape status is established by positioning the
landscape units somewhere along a continuum
defined by an indicator. The condition of a landscape
is judged on the basis of the indicator value — nom-
inal, subnominal, etc. The threshold values are not
21
-------
Landscape Unit
Pattern Unit
Patch Unit
Measurement Unit
Rgure 4.2 Scalable landscape units
22
-------
known for most indicators, but Section 3 described
generally how such thresholds could be determined.
This is an important research question.
Another important question is how to determine a
minimum set of condition indicators to describe
overall landscape status. Ideally, the set would
address all societal values but would hot be redun-
dant. The statistical properties of each would be
well-known, and it would be beneficial if they were
estimable from satellite images. The set would
include integrative measures, so that landscape
change would be detectable no matter what the
causal agent was. It would be undesirable to include
"diagnostic" indicators in this set.
Landscape status can be quantified in terms of the
patterns of land cover (type, distribution, dominance,
and change) and displayed via a G.I.S. Patterns of
land cover play a major role in changing the config-
uration of landscapes at a regional scale. If land
cover patterns change, it is likely that many ecologi-
cal processes will be affected. Areas of predomi-
nantly natural land cover are likely to be in better
overall ecological condition than highly-modified
landscapes. Areas undergoing dramatic changes in
land cover patterns are likely to contain landscapes
whose ecological conditions are changing the fastest.
Other indicators are also being considered for quan-
tifying landscape status (see Section 3).
4.2.4 Data Design Issues
Ideally, status and trends reports will be based on
indicators which are measured on satellite images.
Satellites provide frequent, full coverage of all land-
scapes units. It is a data design issue whether or not
to utilize the available full coverage information, as
opposed to a probability sample of that information.
Flexibility is retained if all landscapes are measured.
If landscapes are sampled, costs are lower but infor-
mation and flexibility are lost. Full-coverage of all
landscape units has a crucial and deciding advantage
over a sampling approach, and that is the flexibility
to reaggregate data in order to assess status and
trends for different configurations of landscapes.
Because landscapes do not have a fixed definition for
all values or indicators, a separate sampling plan
would have to be devised for each situation. Another
benefit is that landscape information will be easier to
combine with the measurements made by all EMAP-
Resource Groups, each of which essentially collects
measurements in different sets or parts of landscapes.
One alternative for sampling is multi-stage sam-
pling. An example of multi-stage sampling is to first
select a sample of arbitrary polygons (e.g., Landsat-
TM or Landsat-MSS scenes) and then to select a sec-
ond-stage sample of landscape units within the first-
stage sample units (say, by a point grid or by list sam-
pling).
If landscapes units are defined as continuous or
spatially extensive resources, then sampling could
associate indicator values with particular points on a
map. A calculation window for an indicator would
be repeatedly placed within a well-defined geograph-
ic area and the indicator value for a given placement
would be assigned to the center point of the window.
The calculation window functions as a "support
region" for the arbitrary point hi space.
A calculation window can be randomly placed
many times to derive an empirical frequency distrib-
ution for the landscape unit, or a systematic grid of
window placements can be used to ensure a good
spatial distribution of sample points. This procedure
can be applied to entire reporting regions, to natural
regions within reporting regions, or to samples with-
in very large landscape units. The resulting "surface"
of points, can then be aggregated by natural region or
reporting region as appropriate for different assess-
ment questions. Re-sampling methods can be used to
develop variance estimates and confidence intervals.
The frequency of image acquisition and process-
ing is a data design issue. The trade offs involve the
degree of image classification, the resolution of dif-
ferent available sensors, and the expected rates of
landscape change.
EMAP-L makes the assumption that Landsat
imagery for the entire U.S. will be available on
approximately 10-year cycles; these data will form
the basis for decadal status, change, and trends
assessments. It is likely, however, that significant
landscape changes will occur more frequently than
on 10-year increments.
EMAP-L will evaluate use of annual AVHRR and
other GIS compatible data to identify areas potential-
ly undergoing significant landscape changes. The
1.1 km2 spatial resolution of the AVHRR will not
permit detailed land cover analysis (Gervin et al.
1985, Loveland et al. 1991) but may be sufficient to
detect major changes resulting from human or natur-
al disturbances. Table 4.1 provides an initial list of
23
-------
Tabla 4.1 Federal Databases Potentially Useful in Identifying Landscape Chanqes
Agency
Database
GIS
Trends
Freq.
Description
DOI.USGS
USDA.ERS
USDA.SCS
USDA.FS
USDA.FS
DOC.Census
AVHRR
MLU
NRI
Yes
No
Yes
FIDO ?
WFS ?
CENDATA Yes
Yes
Yes
Yes
Yes
Yes
5yrs
5yrs
1 yrs
1 yrs
10 yrs
1.1 km data w/ visible & IR channel.
Major Land Uses of the U.S. Provides trends in 11 major
land use classes.
National Resources Inventory. Provides data on status,
trends and condition of soil and water resources, including
soil loss estimated using USLE.
Forest Insect and Disease Conditions. Trends in forest insect
and disease conditions across all forest ownership classes.
Wildland Fire Statistics. Data collected on number and area
burned on public and private lands.
CENDATA, decennial census data. An online data source of
demographic and socioeconomic information. Also include
TIGER line files and TIGER geographic reference products.
GIS compatible data that could be used in combina-
tion with AVHRR to conduct Step 2 change detection
assessments.
For example, graphical analysis of USDA Major
Land Uses (MLU) data (Figure 4.3) shows a sub-
stantial decline in Pennsylvania forest cover since the
middle 1970s, while the amount of forest cover in
Mississippi has remained constant (Both states show
a dramatic increase between 1954 and 1964). In this
example, comparison of MLU data in Pennsylvania
and Mississippi suggests that change detection at
shorter than 10-year intervals would be undertaken in
Pennsylvania.
Combining the MLU data with Bureau of Census
data would suggest that the forest cover loss is in the
southeastern portion of Pennsylvania since the coun-
ties in this portion of the state have experienced pop-
ulation increases greater than 10 percent (U.S. Dept.
of Commerce 1983). Land cover change data from
SCS and NRI could then be examined to 'Verify" the
census data (NRI provides 5-year assessments of
land cover proportion that are statistically reliable at
USGS, 8-digit (small) watersheds). Finally, exami-
nation of temporal AVHRR data could be used to fur-
ther confirm that significant land cover change is
occurring in southeastern Pennsylvania and perhaps
more precisely locate the change. This example is
illustrative of the "weight of evidence" approach that
will be used to monitor land cover dynamics between
10-year Landsat-based land cover detection (Step 2).
The data sources listed in Table 4.1 can be exam-
hied quickly across the entire United States. For
areas determined to have significant landscape
change, EMAP-L would then acquire more detailed
imagery (e.g, Landsat-MSS orLandscape-TM). The
degree of change would then be evaluated to deter-
mine if more detailed evaluation is warranted (see
Step 3 discussion).
4.3 ASSESSMENT APPROACH
, Briefly, an application of the three-step process
(as described in Section 4.1) is as follows. An initial
land cover data set is used to determine landscape
condition in terms of selected indicators (Step 1). A
status report is prepared by using standard statistical
methods as applied to those indicators. After a suit-
able time increment (10 years), new spectral data are
combined with the original spectral data to identify
changed measurement units (pixels) (Step 2). The
changed pixels are then inserted into the original land
cover map. Re-calculated values of indicators for
this updated land cover type map are then used to
update status and trends assessments of landscape
condition. Further analysis of the change determines
whether or not Step 3 is initiated.
The purpose of this section is to fill in some of the
details on how the three-step approach would be
implemented and to provide an example.
24
-------
18000'
17000-
Forest
Area
(1,000 Acres)
16000-
15000-
MS
Redassification
of Forest?
i i i i i r r I 1 I
1945 49 54 59 64 69 74 78 82 87
Combine with Census data at county level would suggest focus on south central and
southeastern counties where population increased more than 10 percent from 1970 to 1980
Figure 4.3 Change in forest area in Pennsylvania
25
-------
4.3.1 Status
Status and trends reports will be prepared based
on the initial land cover information and every 10
years after that. Landscape pattern indicators are an
appropriate starting point for describing status and
trends analyses. Numerous studies have been con-
ducted to develop methods to quantify landscape pat-
tern (Krummel et al. 1987). Information theory mea-
sures (O'Neill et al. 1988a; Li and Reynolds in press)
and fractal dimension (Milne 1992) summarize basic
landscape type and pattern. Studies have shown that
individual indices can capture specific aspects of
landscape pattern.
As a result, there are candidate pattern indicators
available to evaluate in a national landscape monitor-
ing program. For example, dominance (O'Neill et al.
1988a) is an information theoretic index that indi-
cates the extent to which the landscape is dominated
by a single land cover type. The indicator, 0 < D < 1,
is given by the following equation.
D = 1 -
(-Pkln(Pk))/ln(n)]
(1)
where 0 < Pk < 1 is the proportion of land cover type
k, and n is the total number of cover types on the
landscape.
Empirical studies confirm that the fractal dimen-
sion (F) of patches indicates the extent of human
reshaping of landscape structure (Krummel et al.
1987; O'Neill et al. 1988a). Humans create simple
landscape patterns; nature creates complex patterns.
The fractal dimension index is calculated by regress-
ing the log of the patch perimeter against the log of
the patch area for each patch on the landscape. The
index equals twice the slope of the regression line.
Patches of four or fewer pixels are excluded because
resolution problems distort their true shape.
Contagion expresses the probability that land
cover is more "clumped" than the random expecta-
tion (O'Neill et al. 1988a; Li and Reynolds In Press):
The index, 0 < C < 1, is given by equation 2.
C = 1 - C£, 2j (-Py ln(P.p)/(2 ln(n)] (2)
where P,. is the probability that a pixel of cover type
i is adjacent to type j.
Together, this set of indices captures fundamental
aspects of pattern that might influence ecological
processes. Significant changes in these indices indi-
cate that an observed change in the landscape may
result in significant alterations in the quality of the
environment. Experience with dominance, conta-
gion, and fractal dimension in a variety of real-world
settings will undoubtedly highlight aspects of pattern
that are not captured by these indicators.
Cumulative distribution functions (CDFs) are a
convenient way to display the indicator values
obtained for a population of landscape units (see
Figure 4.4). Briefly, a CDF shows the proportion of
the population which has indicator values less than
some nominal value. A CDF is a useful interpretive
device if the nominal value has some meaning, such
as a threshold value separating "good" and "bad"
landscape condition for that indicator. In many cases,
these threshold values can only be arrived at through
research or other experience. Constructing a CDF is
trivial if all landscape units are measured, and it can
be estimated by using standard procedures (Overton
et al. 1990) if only a sample of units are measured. A
time series of CDFs can be displayed as a way of
showing trends.
Indicator values can also be displayed in a map
format (via GIS) in order to show the geographic dis-
tribution of different landscape conditions (Figure
4.4). Landscape units can be color-coded, for exam-
ple, according to the calculated value of an indicator
for that unit. Such a display makes it easier to dis-
cern regional spatial patterns of landscape condi-
tions.
Landscape status will be updated by calculating
land cover change using satellite data. Prior to the
advent of satellite technology, the potential for land
cover change detection using remote sensor data was
noted by Shepard (1964); but, at the time, he also
acknowledged that change detection had an undeter-
minable omission error rate because of the lack of
objectivity. By eliminating the need for human
detection of change (i.e., visual comparison of tl and
t2 air photos), use of satellite or other digital tech-
nology adds a degree of objectivity to such studies
that can not be achieved from aerial photography.
Land cover change using satellite data has been
widely studied (Weismiller et al. 1977, Stauffer and
McKinney 1978, Friedman and Angelici 1979,
Gordon 1980, Byrne et al. 1980, Robinove et al.
26
-------
Time 1
Landscape Contagion Values
0.05-0.15
0.16-0.30
0.15
0.30 0.45
Contagion
0.60
0.15 0.30 0.45
Contagion
0.60
Figure 4.4 Example assessment of landscape status and trends
27
-------
Time 1
Change from Time 1 to Time 2
Landscape Contagion Values
0.05-0.15
0.16-0.30
0.31-0.45
0.46 - 0.60
Time 2
Change in Landscape Contagion Values
No Change
Increase in Contagion
Decrease in Contagion
Step 3 Analysis: watersheds showing
decrease in contagion
Rgure 4.4 Example assessment of landscape status and trends (cont'd.)
28
-------
1981, Jensen 1981, 1983, 1986, Howarth and
Wickware 1981, Jensen et al. 1987, Fung and
LeDrew 1987,1988, Wickham 1988, A. Milne 1988,
Singh 1989, Fung 1990, Hall et al. 1991a, Mouat et
al. 1993). Singh (1989) reviewed nine changed detec-
tion techniques: image differencing, image regres-
sion, image ratioing, vegetation index differencing,
principal components analysis, post classification
techniques, direct multi-date classification, change
vector analysis, and background subtraction (see also
Jensen 1981, 1983, A. Milne 1988 for technique
reviews). We seek a technique that requires minimal
data processing, is sensitive to land cover changes,
and has been widely studied.
Image differencing is a straightforward change
detection technique that has been widely studied
(Weismiller et al. 1977, Stauffer and McKinney
1978, Howarth and Wickware 1981, Jensen 1981,
Nelson 1983, Wickham 1988, A. Mime 1988, Fung
and LeDrew 1988, Fung 1990) and applied in a vari-
ety of geographic settings (Singh 1989). Image dif-
ferencing has the property that spectral differences
between tt and t2 have a near-normal distribution,
where land cover change lies at the tails of the distri-
bution (Stauffer and McKinney 1978). The differ-
ence values that depart only slightly from the mean
are due more to differences in atmospheric condi-
tions, sensor calibration, illumination, soil moisture,
and registration error (Singh 1989). The difference
procedure uses the equation (Jensen 1981):
Ax = x(l)..k - x(2)ijk+c
(3)
where Ax is the change in pixel value, x(l) and x(2)
are pixel values at tx and t2, respectively, i and j are
the pixel row and column position, k is wavelength,
and c is a constant (added to keep the distribution of
change values between 0 and 255).
Where possible, EMAP-L will use land cover
change detection data from the North American
Landscape Characterization Program (NALC) to
update landscape status. NALC recently completed
a review of nine change detection algorithms, and,
not surprisingly, found that image differencing was
the most effective technique for detecting land cover
change (Elvidge et al. 1993).
The most accurate method was image differenc-
ing of all four Landsat MSS bands using a radiomet-
ric correction technique called Automatic Controlled
Scattergram Regression (ACSR). ACSR is a tech-
nique first developed by Hall et al. (1991b) that has
been automated by Elvidge et al. (1993). ACSR uses
a set a bright (e.g., beaches) and dark (e.g., clear
water bodies) pixels that have not changed between
the two dates of satellite acquisition to adjust the
radiometry of one scene (the subject) to the radiom-
etry of the other (reference) scene. The radiometric
adjustment is accomplished using regression tech-
niques. Image differencing follows radiometric
adjustment.
A second technique that was nearly equal to
image differencing using ACSR radiometric adjust-
ment was image differencing of Normalized
Vegetation Index (NDVI) ratios. NDVI ratios
(Tucker 1979) are created by dividing the difference
of the near-infrared (MR) from the red wavelength
by their sum. Detecting land cover change by image
differencing of vegetation indices (e.g., NDVI,
NIR/red ratio) carries two advantages. First, vegeta-
tion indices have been shown to distinguish vegeta-
tion cover from other earth surface features (e.g.,
water, soil) (Knipling 1970, Tucker 1979). Second,
differencing just vegetation indices eliminates the
cost analyzing four Landsat MSS or six TM bands of
change data. Also, change from tj to t2 can be cate-
gorized into two components of coarse estimates of
fragmentation using difference data derived from
vegetation indices. A decrease in the ratio of NDVI
value over time for a given landscape unit provides a
coarse estimate of fragmentation. An increase hi the
ratio of NDVI values over time provides a coarse
estimate of connectivity.
A third technique proposed by NALC is post-
classification subtraction. This technique relies on
complete land cover classification of both tx and t2
dates and then compares the results. NALC propos-
es to use this technique when same season satellite
data can not be acquired for tj and t2. This case is
expected to be rare (Elvidge et al. 1993).
Once the change or difference image has been
created (excluding the post-classification technique),
the question of what value constitutes land cover
change still remains. Distinction of change from no
change is typically defined as a threshold value
(Singh 1989) that is some standard deviation unit
from the mean (Stauffer and McKinney 1978,
Ingram et al. 1981, Singh 1986, Fung and LeDrew
29
-------
1988, Quarmby and Cushine 1989, Fung 1990). This
is illustrated in Figure 4.5 as the area of black hatch-
ing. Fung and LeDrew (1988) found that 0.90 and
1.00 were optimal thresholds for Landsat MSS bands
2 and 4, respectively. A 1.00 threshold was found to
be optimal by Nelson (1983), Fung (1990), and
Quarmby et al. (1987). NALAC is proposing to exam-
ine 0.250, 0.500, and 1.00 change/no change thresh-
olds. However, it is unclear at this tune to what extent
NALC will conduct field accuracy assessments of
their change data on a routine basis. Woodwell (1983)
recommends an interactive visual analysis because it
minimizes incorporating trivial land cover changes,
such as those due to crop rotations. In addition, this
method helps eliminate identifying areas as change
that are the result of anomalies such as cloud cover in
one data set
Woodwell's visual analysis could be set up on a
computer screen as a block of six image sets split into
three pairs. Each image pair would display "before"
and "after" data, and the change identified by the dif-
ferent thresholds (e.g., 0.25, 0.50, 1.0) could be tog-
gled on and off. Visual inspection of the change
results could be compared to air photos and given an
accuracy score using the equation (Cohen I960,
Bishop etal. 1975):
K =
(4)
where K is the Kappa coefficient, M is the total
number of observations, xr are the diagonal elements,
and xi and xj are the off diagonal elements of the 2x2
change/no change matrix shown in Table 4.2. The
threshold with the highest Kappa coefficient (K)
would be used for land cover change detection and
subsequently to update landscape pattern. This
method is proposed for accuracy assessment if the
NALC program does not routinely supply accuracy
estimates of their change detection results.
Table 4.2 Change Detection Matrix
Satellite
Control
Change
Change
No Change
No Change
20 5
5 20
The overall accuracy rates of previous change
detection studies are about 80 percent (Table 4.3).
This is the basis for EMAP-L's assumption that an 80
percent accuracy rate in land cover change detection
can be achieved. However, it may be possible to do
better. Turner (1987a) noted that land cover change is
contagiously distributed. This covariance was not
investigated in any of the change detection studies
cited herein. Because EMAP-L will be conducting
change detection over large geographic areas, it is pos-
sible that EMAP-L might capture some of the omitted
change thereby increasing change detection accuracy.
EMAP-L intends to investigate the phenomenon of
contagion of land cover change to improve change
detection accuracy.
Table 4.3 Reported Accuracy Assessments Of Change
Detection Studies
Author(s)
Fung 1990
Quarmby &
Cushine 1989
Jenson 1981
Fung & LeDrew
Ingram
etal. 1981
Weismiller
etal. 1977
Nelson 1983
Location
Ontario
Canada
England
Denver CO
Ontario
Canada
Denver CO
Matagordo
Harrisburg,
PA
Data Type
Landsat TM
SPOT HRV
Landsat M
Landsat MSS
Landsat MSS
Landsat MSS
Landsat MSS
Landsat MSS
Overall
Bands
DI3
DI4
DI5
HRV 3-
TM4
DI5
DI7
DI5
DI4
DI5
DI6
DI7
Dl
DI4
DI5
DI6
DI7
Accuracy
W
83.3
82.2
79.6
NR(!)
77.0
80.8
77.3
84.9
83.8
81.5
72.4
NR
83.3
80.6
86.8
84.2
Dl = difference image
(!)= authors reported 15% omission error
Landsat MSS
Band 4:500- BOOnm
Band 5:600- 700nm
Band 6:700- SOOnm
Band 7:800-1100nm
Landsat TM
Bandl: 450-520nm
Band 2: 520-600nm
Band 3: 630-690nm
Band 4: 760-900nm
Band 5:1550-1750nm
Band7:2080-2350nm
SPOT HRV
Band1:500-590nm
Band2:610-680nm
Band 3:790-890nm
4.3.2 Analyzing Landscape Change
Once the changes of individual pixels have been
determined, two types of analyses will be carried out.
The first analysis is simply to update the status and
trends of landscape condition, as described above.
30
-------
Time 1
Landsat Scene
Time 2
Landsat Scene
Register
Scenes
Time 1
Subset
RANGE
Time 1
(BAND 5)
Pixel by Pixel
Subtraction
Time 1 - Time 2
+ Constant (128)
10
Minus
(10-40)+ 128 = 98
RESIDENTIAL
PIXEL OF CHANGE
255
Figure 4.5 Image Substraction for Land Cover Change Detection
31
-------
The second analysis is to determine whether or not
the observed changes warrant in-depth evaluation of
associations between observed landscape status and
trends with environmental stressors. The second
analysis requires additional explanation here.
The types and magnitude of change identified in
Step 2 may trigger more in-depth assessments to be
carried out in Step 3. A number of research criteria
will be investigated for possible use, including the
concept of "phase" change in landscape pattern state
space. Despite the success of individual pattern indi-
cators, no single index captures all of the relevant
aspects of pattern. Therefore, a more complex
method, involving at least two or three dimensions, is
needed to classify pattern types. Each axis is nor-
malized to unit length so that all indicators make
equal contributions. The landscape unit is then rep-
resented by a point in this abstract state space.
In Figure 4.6, an undisturbed grassland landscape
has high values for three indicators. A similar land-
scape, when converted to agricultural uses, has high
dominance and contagion but low fractal dimension.
Thus, the position of a landscape hi the 3-dimension-
al state space illustrated in Figure 4.6 will immedi-
ately orient the viewer to the type of landscape pat-
tern in the subject landscape unit. The approach also
permits quantitative description of changes in a given
landscape unit over time. Changes hi land cover pro-
duce changes in indicator values and corresponding
movement of the landscape unit hi 3-dimensional
state space. The distance moved is a measure of the
amount of change in a landscape. This is calculated
by the following equation:
(5)
At some (as yet unknown) magnitude, this shift will
represent a phase change hi the landscape.
Another way to summarize changes hi land cover
between two times is through the use of transition
matrices (see Table 2.1). The diagonal elements rep-
resent the degree to which land cover types do not
change over time. If the rows and columns are cho-
sen carefully, the transition matrix also provides a
picture of disturbance regimes and succession! The
upper triangular elements can represent succession
and the lower triangular elements can represent dis-
turbance (Hall et al. 1991a).
In some cases, an indicator value is expected to
change when the underlying sample unit changes hi
some important way. For example, as large natural
forest patches are fragmented into small, regular
patches by urban development, both the shape and
size of the associated landscape sample units vary
predictably. This correlation is important, especially
because one of the two ways proposed for defining
landscape objects (i.e., as LPTs) yield landscape pop-
ulations that may change over time. Comparisons of
some indicators over time or space may have to be
conditioned upon the size of the underlying land-
scape units, for example, by covariance analysis.
4.3.3 In-depth assessment of Landscape
Condition and Associations with Stressors
Change detection analyses will identify areas
undergoing a level of change hi landscape pattern
that will be likely correlated with changes in condi-
tions of landscape values. In these areas, we will
conduct in-depth analyses of landscape condition.
In-depth analysis of landscape condition refers to
greater specificity hi the calculation of landscape
indicators, and this often requires use of ancillary
data. For example, within areas undergoing rapid
change, we may calculate landscape pattern indica-
tors on the scale of a "patch" to evaluate within-patch
composition and structure relative to landscape sus-
tainability. Landscape indicator measurements at
this scale would require use of ancillary data such as
aerial photography.
We also propose to evaluate associations between
landscapes undergoing significant change with
anthropogenic and natural stressors. We anticipate
using ancillary data on stressors, including U.S.
Census Bureau data, NRI data, and USGS digital
data on roads (to name a few) to determine these
associations. Most of these data will come hi the
form of GIS coverages.
We also will analyze landscape condition as it
relates to conditions of other ecological resources
imbedded within landscapes, such as Federally-listed
species, breeding birds, and those of EMAP-
Resource Groups (e.g., streams). This will require
use of ancillary data (e.g., U.S. Breeding Bird
Survey) and data generated by EMAP Resource
Groups.
The primary differences between an in-depth
analysis and a status report are increased specificity
of landscape condition indicators and the introduc-
tion of ancillary data for more in-depth analysis.
32
-------
Clumped
Clumped
Figure 4.6 Three-dimensional landscape indicator space
33
-------
-------
5.0 IMPLEMENTATION OF EMAP-LANDSCAPES
EMAP-L proposes to implement its program
simultaneously along two lines: (1) a series of pilots
to address research and development issues and (2)
determination of landscape status and trends within
pilot areas for those indicators ready for implementa-
tion. Additionally, EMAP-L will conduct workshops
and focused research on landscape values, assess-
ment questions, and conceptual models.
EMAP-L is proposing two general pilot areas:
one hi the east and one in the west. The eastern pilot
will be hi the Mid-Atlantic Region of the United
States. In FY94, work will be initiated in the
Chesapeake Bay watershed part of this region; work
will be expanded to the remainder of the area in
FY95. The western pilot will be initiated in FY95.
The location of the western pilot has not been deter-
mined although the Pacific Northwest, Rio Grande
Basin, and Colorado Plateau are the leading candi-
dates. The following criteria have been and will con-
tinue to be used for selection of pilot areas: (1)
extensive existing remote sensing and GIS databases;
(2) diverse or different landscape mosaics and condi-
tions; (3) regional management focus and landscape
emphasis with a number of participating organiza-
tions; (4) significant potential for collaboration with
other EMAP Resource Groups and other Federal pro-
grams (e.g., GAP).
5.1 RESEARCH AND DEVELOPMENT
Research and development will address issues
that must be resolved in order to fully implement the
three-step landscape monitoring approach. These
include (1) refining landscape values upon which the
program will be based-; (2) developing additional
assessment questions related to each landscape
value; (3) refining landscape conceptual models
related to each value; (4) determining which land-
scape units and scales are appropriate for evaluating
landscape values and associated assessment ques-
tions; (5) developing and testing landscape indicators
and associated analysis techniques; and (6) develop-
ing and testing the three-step monitoring approach
including key components in each step. Table 5.1
summarizes a preliminary list of research questions
to be addressed by EMAP-L. Findings hi each of
these areas will be used to modify and implement
key components of the three-step monitoring
approach.
5.1.1 Landscape Values/Assessment
Questions/Conceptual Models
Our aim is to identify societal values that relate,
in part, to status and trends hi the condition of land-
scapes. Biotic integrity and diversity, watershed
integrity, and landscape sustainability and resiliency
are the primary landscape values to be addressed by
the landscape monitoring program. Research is
needed to refine the landscape values and to list
assessment questions relative to each value which
help to define the scope of the monitoring. We pro-
pose to conduct an extensive literature review of
societal values related to landscapes and to hold
workshops and meetings with scientists, managers,
and the public representing different regions of the
United States to refine landscape values and to devel-
op assessment questions. Additionally, we propose to
compile existing conceptual models relevant to each
landscape value and to refine models, as necessary.
5.1.2 Landscape Indicators
A key component of the landscape monitoring
approach is the set of landscape indicators that asso-
ciate landscape patterns to landscape values. We
35
-------
Table 5.1 Preliminary list of research and development questions to be addressed by EMAP-Landscapes
Landscape Values and Assessment Questions
1. What are the primary hypotheses of association between landscape values and stressors and how do they vary by region?
2. What variation is there in the application of these landscape values to different regions of the country?
3. What is the specific set of assessment questions relative to landscape values?
Landscape Conceptual Models and Indicators
4. What models of landscape function, structure, and composition best describe landscape values?
5. What fundamental spatial and temporal scales are revealed by these models? *
6. Which of the some 40 existing landscape indicators are independent and sensitive enough to describe status and trends of landscapes?
7. What are the levels of sensitivity (e.g., related to status and trends) of candidate landscape indicators? Do these sensitivities vary by region
and scale?
8. To what degree do differences in classification of imagery affect landscape indicator performance? How do different levels of classification
accuracy affect indicator performance?
9. What accuracy assessment sampling scheme is necessary to achieve a minimum of 80% accuracy for each ecological resource class used in
the landscape analysis?
10. What spatial and temporal scales within each general landscape classification are important in describing status and trends in landscapes?
11. How do decisions about landscape units (e.g., type, size, shape, and scale) affect indicator performance?
Three-Step Monitoring Paradigm
12. Does land cover change detected using AVHRR accurately represent the change that would be detected using Landsat MSS?
13. What density and extent of land cover change detected with Landsat MSS and TM is necessary to trigger Step 2 characterization of land
scape pattern and trends?
14. Do the improved sensor characteristics of Landsat TM as compared to MSS provide significantly more accurate land cover change detection
results for Anderson (1976) Level I land cover categories?
15. What is the minimum size of landscape unit characterization (watersheds and landscape pattern types) relative to the density and extent of
land cover change observed in Step 1 ? And does this size vary across the continental United States?
16. How well do selected landscape indicators relate to landscape values (e.g., watershed integrity), and can the variance in these indicators be
used to construct nominal/subnominal boundaries for different landscape patterns (Step 3)?
17. What are the relative advantages of calculating landscape indicators from wall-to-wall imagery versus "sub-samples" of images?
have identified a preliminary list of landscape indi-
cators (Table 5.2) and propose to compare these to
landscape models developed for each landscape
value; this will help us eliminate some of candidate
indicators and identify others.
As the list of potential indicators grows, they will
be subjected to a series of quality tests. Existing lit-
erature (e.g., O'Neill et al. 1988a) indicates that it is
necessary to determine the range for each indicator
of existing data for a wide variety of landscapes. The
indicator may have a potential range from 1.0 to 2.0.
But if diverse landscapes actually range from 1.3 to
1.35, the indicator may be of little use hi discrimi-
nating patterns of real landscapes.
We will test for sensitivity by randomly varying
pixels on existing landscape data sets; sensitivity
tests will be conducted hi pilot areas. A sensitive
index will change with few pixel changes, whereas
an insensitive index will change only slowly.
Insensitive indicators may be useful for general clas-
sification purposes but will not be relied upon for
trend analysis. On the other hand, an overly sensitive
indicator may jump to the opposite end of its range
with very few pixel changes and may be useless for
overall assessment purposes.
Landscape indicators will be tested for then-
response (or sensitivity) to grain size and the number
of classes. Indicator responses as a function of the
number of classes are important because classified
imagery is likely to have a wide range hi the number
of classes (e.g., 8Lclass NALC versus 25- to 30-class
GAP analysis data).
The indicators will be tested for independence by
calculating multiple indicators on existing landscape
data and calculating covariances. Experience has
shown that different indicators may, in fact, be cap-
turing the same aspects of pattern; these tests will be
conducted on pilot areas.
The indicators will be subjected to uncertainty
analysis by field tests. The tests will be conducted hi
collaboration with EMAP resource groups and other
programs working hi the field (e.g., U.S. Soil
36
-------
Conservation Service's NRI). The suite of indicators
wffl be calculated ftom imagery and will be tested for
accuracy, sensitivity to change, etc., via ground truth.
These tests will be conducted in pilot areas.
Finally, it is a research issue to determine thresh-
olds of landscape indicator values as they pertain to
landscape condition and each landscape value.
Indicators that pass through these tests will be
implemented within the region hi which they were
tested. It is likely that some implemented landscape
indicators will be dropped in favor of indicators with
greater resolving power; these latter indicators will
come from ongoing research and development activ-
ities within and outside the EMAP-L program.
5.1.3 Three-step Monitoring Approach
Implementing EMAP-L's three-step monitoring
approach involves answering many technical and
methodological questions; these include evaluating
indicators, as well as other issues (see Table 5.1).
EMAP-L will address these issues through pilot stud-
ies and simulations.
We propose to acquire multi-date Landsat-MSS
(early 1970s, 1980s, and 1990s), and Landsat-TM
coverages for each of the pilot areas. One objective
is to evaluate the decision criteria (e.g., the magni-
tude and distribution of change) used to trigger Step
3 analysis. We will test Step 3 association analyses
by comparing landscape indicators hi areas undergo-
ing known changes with data on both natural (e.g.,
climate, fire, floods) and anthropogenic (e.g., land
use type and distribution) stresses.
A second objective is to compare landscape indi-
cator values derived from Landsat-MSS images with
those derived from Landsat-TM. If Landsat-MSS
indicator values are similar to those derived from
Landsat-TM, then there is an opportunity to conduct
a retrospective analysis of landscape status and
change, with a baseline of the early 1970s; these data
will then be comparable to future estimates of land-
scape status and trends derived from Landsat-TM.
The major advantage of using these data is that
EMAP-L could produce an assessment of landscape
status and trends within pilot areas and, potentially,
over most of the United States. A similar Landsat-
MSS database is being used to evaluate land cover
change over the entire Australian continent (Dean
Graetz, CSIRO-Australia, personal communica-
tions).
EMAP-L will evaluate the use of different ecore-
gional approaches (e.g., Bailey, Omernick, and
Kuchler) to characterize landscapes and two methods
of classifying landscapes within ecoregions: by
watersheds and by land cover pattern types
(Wickham and Norton 1994). EMAP-L will also
evaluate the relative advantages and disadvantages of
calculating landscape indicators on a "sub-sample"
of landscapes (e.g., a randomly selected subset of the
total landscapes) versus all landscapes within a
region.
Finally, the hypothesis that annual AVHRR and
other agency data are sufficient to detect landscape
changes between 10-year status and trends reports
must also be tested. The basic concept is to perform
change detection analysis using annual AVHRR data
and other agency data, and to estimate landscape
change and to compare these results to those derived
from Landsat-TM and Landsat-MSS data. The
objective is to estimate the risk of missing a signifi-
cant change in land cover between 10-year resam-
ples.
5.2 IMPLEMENTATION OF LANDSCAPE
MONITORING
The rate of implementation of EMAP-
Landscapes will depend on our success in resolving
key technical issues and on funding scenarios.
Implementation of the program nationally will pro-
ceed by (1) increasing the area of pilot studies to
cover applicable natural regions (e.g., Bailey's ecore-
gions) and (2) adding oh new regions. The schedule
for adding on new regions will be coordinated with
EMAP-LC and EMAP Resource Groups.
Coordination with EMAP Resource Groups will be
important especially when both landscape and
Resource Group level data (e.g., fine scale habitat
characteristics) are needed to address a common
value (e.g., biotic diversity).
We anticipate initiating status and trends assess-
ments in pilot areas for those indicators ready for
implementation (see Table 5.2). As other indicators
pass through a series of sensitivity and uncertainty
analysis, they will be added to the status and trends
assessments.
37
-------
Table 5.2 Status and Utility of Landscape Indicators
Statusfc) Utility(b)
WATERSHED INTEGRITY INDICATORS
Dominance C ST
Fractal dimension C ST
Contagion C ST
Lacunarity A ST
Boston risk AD
Flood Indicator AD
Riparian zones C D
Loss of wetlands C D
Agriculture near water B D
Miles of new roads B D
Amounts of agriculture and urban C D,ST
Watershed/water quality indicator A D
LANDSCAPE STABILITY AND RESILIENCE INDICATORS
Dominance C ST
Fractal dimension C ST
Contagion C ST
Lacunarity A ST
Diffusion rates
Percolation backbone
Percolation thresholds
Miles of roads
Recovery time
Land cover transition matrix
A
B
B
B
A
A
B10T1C INTEGRITY AND DIVERSITY INDICATORS
Dominance C
Fractal dimension C
Contagion C
Lacunarity A
Change of habitat • C
Habitat for endangered species C
Loss of rare land cover C
Corridors between patches B
Amount of Edges C
Edge amount per patch size B
Patch size distribution C
Table 52 Continued.
D
D
D
D
D
D
ST
ST
ST
ST
D.ST
D
D
D,ST
D,ST
D,ST
D.ST
Status(a) Utility(b)
Largest patch B D,ST
Inter-patch distances B D,ST
Linear configurations A D,ST
Actual vs. Potential vegetation B D
Wildlife potential A D,ST
Miles of new roads B D
Diffusion rates A D
Percolation backbone B D
Percolation thresholds B D
Resource utilization scale B D,ST
Scales of pattern A D,ST
Cellular automata A D
Pixel transitions A D
(a) Status categorizes each indicator as:
A * Requiring further conceptual development
B » Requiring testing for feasibility/sensitivity
C M Ready for field tests and implementation
(b) Utility categorizes each indicator for use in:
ST=Status and trends assessments
D m Diagnostic or association assessments
Our ability to produce retrospective change and
trend analysis within pilot areas over the next 3-5
years will depend on results of research on Landsat-
MSS, as well as funding to label Landsat-MSS pro-
vided by NALC.. If both of these conditions are met,
we propose to produce an initial landscape status
assessment (T0) for the early 1970s. A second status
assessment would be produced from the mid-1980s
data, and differences between the early 1970 and
mid-1980 images (tQ - ta) would be used to identify
areas undergoing changes in landscape pattern. We
would then conduct Step 3 assessments in areas
meeting change criteria. We would repeat this
process by comparing mid-1980s and early 1990s
data (tj - t2). By the mid to late 1990s, Landsat-TM
should be available for the entire country. Landsat-
TM would be used for the fourth sample in the series
5.3 COLLABORATIVE EFFORTS
PROPOSED BY EMAP-L
EMAP-L anticipates close collaboration with
EMAP Resource Groups and other monitoring and
assessment programs as described below.
• EMAP and the Biological Diversity Research
Consortium propose to jointly investigate the con-
dition of biological diversity components (e.g.,
habitat condition, species richness) on regional
and national scales. EMAP-L will use assess-
ments of landscape status and trends to help char-
acterize risk to biological diversity components.
EMAP-LC has entered into an agreement with the
USGS National Aquatic Water Quality
Assessment Program, NOAA's Coastwatch
Change Analysis Program, and the U.S. National
Biological Survey's GAP Analysis Program to
acquire and classify wall-to-wall Landsat-TM
imagery for the United States. EMAP-L proposes
to work with this group in obtaining Landsat-TM
data for calculating landscape indicators.
NALC, in collaboration with the USGS EROS
Data Center, is processing Landsat-MSS data for
three time periods in a large portion of the United
States-. NALC is also developing change detec-
tion techniques for use with these data. EMAP-L
38
-------
proposes to collaborate with. NALC and to use
NALC's Landsat-MSS data to begin testing
EMAP-L's monitoring strategy within pilot areas.
EMAP-L and EMAP-LC will work closely on a
number of issues, including development of land-
scape classifications for calculating landscape
indicators and accuracy assessment protocols.
EMAP-L anticipates collaborating with EMAP
Resource Groups to address common societal val-
ues such as biotic diversity within the pilot areas,
and to obtain ground truth of remote sensing data.
5.4 PROJECT MILESTONES/ACTIVITIES
A general schedule of EMAP-L activities and
milestones anticipated over the next 5 years is listed
in Table 5.3. Dates listed for activities and mile-
stones represent EMAP-L's best estimates based on
budget projects. However, dates listed for activities
and milestones are likely to deviate from those listed,
especially in latter years.
Additionally, pilot studies other than those listed
may be initiated in FY94 - 98 due to changes in
EMAP and EPA priorities.
EMAP-L anticipates working closely with EPA
regional offices and EPA's Risk Assessment
Forum on issues relating to characterization and
assessment of regional ecological risks. The
degree of collaboration will depend on mutual
interests and benefits for specific questions.
39
-------
•fc
i
0)
o\
111
CO
I
.-2
(B
-
ill
t
3
i
w
I
'II
si
0)
(0
CO
in
o>
A
.CB
40
-------
fl
1
£
•^
50
siiu
S
sp
§
.9?
tyt • JJ
_£ §1
m gg
t 81
^ !sg£
^ in
^" ' '- ^ ^ ^3§
^* ^r "JJ
I - i i -|
•5
Schedule
CO
ui
o>
A
i
i
8
1
0)
S
•o
CO
w
0)
I
a.
<
UJ
41
-------
SCO
1
CO
CO
PH JO
OC
- B
o s
2 2
'ill
o
\
C 05
.
*- CD
O Q)
t
i
o
I
P
i in
i
t
CB
1
0)
t
I
IP
2
I
(0
1
(0
(0
0)
UJ
0)
0)
m
10
42
-------
Q
O
Q
O
-
.....
ntinued
S-
o
o>
Q.
•o
i
in
0)
43
-------
Q
•55
00
OS
ON
Q
00
1
t
B S
PT
-------
Q
S5
O
oa
F:
o
'42
<3
(O
I
£
08
,
' - " J'fli
^*cn
m S*
''".." '"'"' 111
Sfl '^ «D §
- -" ' fcS "- -|Sl
!• 1 §>
s- - - § § '
...... ,| - ...s g
TO
0)
o
m
CO
U)
45
-------
-------
6.0 LITERATURE CITED
Ales, R.F., A. Martin, F. Ortega, and E.E. Ales.
1992. Recent changes in landscape structure
and function in a mediterranean region of SW
Spain (1950-1984). Landscape Ecol. 7:3-18.
Anonymous. 1980. An approach to water
resources evaluation of non-point silvicultural
sources (WRENS). A Procedural Handbook.
Forest Service, Washington, DC EPA
600-8-80-012.
Bailey, R.G. 1991. Design of ecological net-
works for monitoring global change. Environ.
Conserv. 18:173-175.
Bailey, R.G., M.E. Jensen, D.T. Cleland, and P.S.
Bourgeron. 1993. Design and Use of
Ecological Mapping Units. Vol n Ecosystem
Management: Principles and Applications.
Berry, J.K., and J.K. Sailor. 1987. UseofGIS
for storm runoff prediction from small urban
watersheds. Environmental Management
11:21-27.
Bierregard, R.O. 1990. Avian communities in
the understory of Amazonian forest frag-
ments. Pp. 333-343, in A. Keast (ed.)
Biogeography and ecology of forest bird com-
munities. SPB Academic Publishing, The
Hague, The Netherlands.
Bourgeron, P.S. and M.E. Jensen. 1993. An
overview of ecological principles for ecosys-
tem management. Pp. 49-60, in Jensen, M.E.,
and P.S. Bourgeron (eds.), Eastside Forest
Ecosystem Health Assessment, Vol II,
Ecosystem Management: Principles and
Applications. U.S.D.A. Forest Service,
Missoula, Montana.
Byrne, G.F., P.F. Crapper, and K.K. Mayo. 1980.
Monitoring land-cover change by principal
component analysis of multitemporal Landsat
Data. Remote Sensing of Environment,
10:175-184.
Comins, H.N., and LR. Noble. 1985. Dispersal,
variability, and transient niches; species coex-
istence in a uniformly variable environment.
Am.'Nat. 126:706-723.
Correll, D.L. 1991. Human impact on the func-
tioning of landscape boundaries. Pp. 90-109,
in: The Role of Landscape Boundaries in the
Management and Restoration of Changing
Environments, M.M. Holland, P.G. Risser and
R.J. Naiman (eds.), Chapman and Hall, NY.
Danielson, B.J. 1992. Habitat selection, inter-
specific interactions and landscape composi-
tion. Evol. Ecol. 6:399-411.
DelRegno, K.J., and S.F. Atkinson 1988.
Nonpoint pollution and watershed manage-
ment: remote sensing and GIS approach.
Lake and Reservoir Management 4:17-25.
Donigian, A.D., D.C. Beyerlein, H.H. Davis, and
N.H. Crawford. 1977. Agricultural Runoff
Management (ARM) Model Version II:
Refinement and testing. USEPA, Athens, GA
EPA 600/3-77-098.
Elvidge, D.D., D. Yuan, and R. Weerackoon.
1993. Change detection methods. Report
Contract #CR816826-02. U.S. Environmental
Protection Agency. Environmental Monitoring
Systems Laboratory, Las Vegas, Nevada.
Environmental Protection Agency (EPA). 1993.
The watershed protection approach: annual
report 1992. U.S. EPA 840-5-93-001.
Everett, R., P. Hessburg, M. Jensen, and B.
Bormann. 1993. Eastside Forest Ecosystem
Health Assessment. Vol I Executive Summary.
Flather, C.H., S.J, Brady, and D.B. Inkley. 1992.
Regional habitat appraisals of wildlife com-
munities: a landscape-level evaluation of a
resource planning model using avian distribu-
tion data. Landscape Ecol. 7:137-147.
47
-------
Forman, R.T.T. 1990. Ecologically sustainable
landscapes: the role of spatial configuration.
Pp. 261-278, in Forman, R.T.T., and I.S.
Zomeveld (eds.), Changing landscapes: an
ecological perspective.
Forman, R.T.T., and M. Godron. 1986.
Landscape Ecology. Wiley and Sons, NY.
Franklin, J.F., K. Gromack, W. Denison, et al.
1981. Ecological characteristics of old-
growth Douglas-fir forests. USDA-Forest
Service General Technical Report PNW-118.
Pacific Northwest Forest and Range
Experiment Station, Portland, Oregon.
Franklin, J.F., and R.T.T. Forman. 1987.
Creating landscape patterns by forest cutting:
ecological consequences and principles.
Landscape Ecology 1:5-18.
Freidman, S.Z., and G.L. Angelici. 1979. The
detection of urban expansion from Landsat
imagery. Remote Sensing Quarterly, 1:58-78.
Fung, T. 1990. Assessment of TM imagery for
land cover change detection. IEEE
Transactions on Geoscience and Remote
Sensing, 28(4):681-684.
Fung, T., and E. LeDrew. 1987. Application of
principal components analysis to change
detection. Photogrammetric Engineering and
Remote Sensing, 53(12):1649-1658.
Fung, T., and E. LeDrew. 1988. The determina-
tion of optimal threshold levels for change
detection using various accuracy indices.
Photogrammetric Engineering and Remote
Sensing, 54(10): 1449-1454.
Gardner, R.H., B.T. Milne, M.G. Turner, and
R.V. O'Neill. 1987. Neutral models for the
analysis of broad-scale landscape pattern.
Landscape Ecology, 1: 19-28.
Gardner, R.H., R.V. O'Neill, M.G. Turner, and
V.H. Dale. 1989. Quantifying scale-depen-
dent effects of animal movement with simple
percolation models. Landscape Ecology,
3:217-227.
Gardner, R.H., M.G. Turner, R.V. O'Neill, and S.
Lavorel. 1991. Simulation of the scale-depen-
dent effects of landscape boundaries on
species persistence and dispersal. Pp 76-89,
in The Role of Landscape Boundaries in the
Management and Restoration of Changing
Environments, M.M. Holland, P.G. Risser and
RJ. Naiman (eds.), Chapman and Hall, New
York.
Gardner, R.H., V.H. Dale, R.V. p'Neill, and
M.G. Turner. 1992. A percolation model of
ecological flows. Pp 259-269, in A.J. Hansen
and F. di Castri (eds.), Landscape Boundaries:
Consequences for Biotic Diversity and
Ecological Flows. Springer-Verlag, New
York.
Gardner, R.H., R.V. O'Neill, and M.G. Turner.
In press. Ecological implications of landscape
fragmentation. Gary Conference IV:
"Humans as Components of Ecosystems,"
Institute of Ecological Studies, Millbrook,
NY, 4-8 May 1991. Springer Verlag, NY.
Geritz, S.A.H., J.A.J. Metz, P.G.L. Klinkhamer,
and T.J. De Jong. 1987. Competition in safe
sites. Theor. Pop. Biol. 33:161-180.
Golley, F.B. 1989. Landscape ecology and bio-
logical conservation. Landscape Ecol.
2:201-202.
Golley, F.B. 1993. Development of Landscape
Ecology and its Relationship to
Environmental Management. Pp. 37-44, in
Jensen, M.E., and P.S. Bourgeron (eds.),
Eastside Forest Ecosystem Health
Assessment, Vol n, Ecosystem Management:
Principles and Applications. U.S.D.A. Forest
Service, Missoula, Montana.
Gordon, S.I. 1980. Utilizing Landsat imagery to
monitor land-use change: a case study in
Ohio. Remote Sensing of Environment,
9:189-196.
Gosselink, J.G., G.P. Shaffer, L.C. Lee, D.M.
Burdick, D.L. Childers, N.C. Leibowitz, S.C.
Hamilton, R. Boumans, D. Cushman, S.
Fields, M. Koch, and J.M. Visser. 1990.
Landscape conservation in a forested wetland
watershed. BioScience 40:588-600.
Graham, R.L., C.T. Hunsaker, R.V. O'Neill and
B.L. Jackson. 1991. Ecological risk assess-
ment at the regional scale. Ecological
Applications 1:196-206.
Gregory, S.V., F.J. Swanson, W.A. McKee, and
K.W. Cummins. 1990. An ecosystem per-
spective of riparian zones. Bioscience
41(8):540-551.
48
-------
Grover, H.D., and H.B. Musick. 1990.
Shrubland encroachment in southern New
Mexico, U.S.A.: an analysis of desertification
processes in the American southwest.
Climatic Change, 17:305-330.
Hall, E, D.B. Botkin, D.E. Strebel, K.D. Woods,
and SJ. Goetz. 1991a. Large-scale patterns of
forest succession as determined by remote
sensing. Ecology, 72(2):628-640.
Hall, EG., D.E. Strebel, J.E. Nickeson, and SJ.
Goetz. 1991b. Radiometric rectification:
toward a common radiometric response
among multidate, multisensor images.
Remote Sensing of Environment, 35:11-27.
Hann, W., M.E. Jensen, and M. Prather. 1993.
Land Management Assessment Using
Hierarchical Principles of Landscape Ecology.
Pp.. 293-306, in Jensen, M.E., and P.S.
Bourgeron (eds.), Eastside Forest Ecosystem
Health Assessment, Vol IE, Ecosystem
Management: Principles and Applications.
U.S.D.A. Forest Service, Missoula, Montana.
Harris, L.D. 1988. Edge effects and conserva-
tion of biotic diversity. Conserv. Biol. 2:330-
332.
Harris, L.D., and Gallagher. 1989. New initia-
tives for wildlife conservation: the need for
movement corridors. Pp. 11-34, in G.
Mackintosh (ed.), Preserving communities
and corridors. Defenders of Wildlife,
Washington, D.C.
Hession, W.C., and V. p. Shanholtz 1988. A GIS
for targeting nonpoint source agricultural pol-
lution. J Soil and Water Conservation
43:264-266.
Holling, C.S. 1973. Resilience and stability of
ecological systems. Annual Review of
Ecology and Systematics. 4:1-23.
Holling, C.S. 1992. Cross-scale morphology,
geometry, and dynamics of ecosystems. Ecol.
Monogr. 62:447-502.
Howarth, P.J., and G.M. Wickware. 1981.
Procedures for change detection using
Landsat digital data. International Journal of
Remote Sensing, 2(3):277-291.
Hunsaker, C.T., D.A. Levine, S.P. rimmins, B.L.
Jackson, and R.V. O'Neill. 1992. Landscape
Characterization for assessing regional water
quality. In Ecological Indicators, In Press.
Ingram, K., E. Kanpp, and J.W. Robinson. 1981.
Change detection technique development for
improved urbanized area delineation.
Technical Memorandum CSC/TM-81/6087,
Comp. Sci. Corp., Silverspring, Maryland,
U.S.A.
Jensen, J.R. 1981. Urban change detection map-
ping using Landsat digital data. The
American Cartographer, 8(2): 127-147
Jensen, J.R. 1983. Urban/suburban land use
analysis. In Colwell, R. N. (Ed.) Manual of
Remote Sensing. American Society of
Remote Sensing, Falls Church, Virginia,
USA, pp. 1571-1666.
Jensen, J.R. 1986. Introductory Digital Image
Processing. Prentice-Hall, Engelwood Cliffs,
New Jersey.
Jensen, J.R., and P.S. Bourgeron (eds.). 1993.
Eastside Forest Ecosystem Health
Assessment, Vol U., Ecosystem Management:
Principles and Applications. U.S.D.A. Forest
Service, Missoula, Montana.
Jensen, J.R., E.W. Ramsey, H.E. Mackey Jr., E.J.
Christensen, and R.R. Sharitz. 1987. Inland
wetland change detection using aircraft MSS
data. Photogrammetric Engineering and
Remote Sensing, 53(5):521-529.
Jensen, M.E., and R. Everett. 1993. An
Overview of Ecosystem Management
Principles. Pp. 7-16, in Jensen, M.E., and P.S.
Bourgeron (eds.), Eastside Forest Ecosystem
Health Assessment, Vol II, Ecosystem
Management: Principles and Applications.
U.S.D.A. Forest Service, Missoula, Montana.
Kareiva, P. 1986. Patchiness, dispersal, and
species interactions: consequences for com-
munities of herbivorous insects. Pages
192-206, in J. Diamond and T.J. Case, eds.
Community ecology. Harper and Row, New
York.
Karr, J.R., and I.J. Schlosser. 1978. Water
resources and the land-water interface.
Science 201:229-233.
Knipling, E.B. 1970. Physical and physiologi-
cal basis for the reflectance of visible and
near-infrared radiation from vegetation.
Remote Sensing of Environment, 1:155-159
49
-------
Knisel, W.G. (ed.) 1980. CREAMS: A field-
scale model for chemicals, runoff, and erosion
from agricultural management systems.
Conservation research report 20. US
Department of Agriculture, Washington DC.
Krummel, J.R., R.H. Gardner, G. Sugihara, R.V.
O'Neill, and P.R. Coleman. 1987. Landscape
patterns in a disturbed environment. Oikos
48:321-324.
Kiichler, A.W. 1964. Potential natural vegeta-
tion of the conterminous United States.
American Geogr. Soc., Spec. Publ., No. 36,
Washington, D.C., 116 pp.
Ledig, F.T. 1992. Human impacts on genetic
diversity in forest ecosystems. Oikos 63:87-
108.
Levin, S.A. 1976. Spatial patterning and the
structure of ecological communities. Lecture
on Mathematics in the Life Science 81:1-35.
Levin, S.A. 1978. Pattern formation in ecologi-
cal communities. Pp. 433-465, in J.H. Steele
(ed.), Spatial Pattern in Plankton
Communities. Plenum, New York.
Levine, D.A. 1992. A Geographic Information
System Approach to Modeling Nutrient and
Sediment Transport. PhD Thesis, School of
Public and Environmental Affairs, Indiana
University.
Levine, D.A., and W.W. Jones 1990. Modeling
phosphorus loading to three Indiana reser-
voirs: a GIS approach. Lake and Reservoir
Management 6:81-91.
Li, H., and J.F. Reynolds. A new contagion
index to quantify patterns of landscapes.
Landscape Ecology, In Press.
Likens, G.E., F.H. Bormann, R.S. Pierce, and
W.A. Reiners. 1978. Recovery of a deforest-
ed ecosystem. Science 199:492-496.
Lillesand, T.M., andR.W. Kiefer. 1987. Remote
Sensing and Image Interpretation. 2nd Ed.
John Wiley & Sons, New York.
Loucks, O. 1970. Evolution of diversity, effi-
ciency, and community stability. American
Zoologist, 10:17-25
Lovejoy, T.E., R.O. Bierregard, and A.B.
Rylands. 1986. Edge and other effects of iso-
lation on Amazon forest fragments. Pp.
275-285, in M.E. Soule (ed.), Conservation
Biology: The science of scarcity and diversity.
Sinauer Assoc., Sunderland, Massachusetts.
Lowrance, R.R., R. Leonard, and J. Sheridan.
1985. Managing riparian ecosystems to con-
trol nonpoint pollution. J Soil and Water
Conservation 40:87-91.
Lowrance, R.R., R.L. Todd, and L.E. Asmussen.
1983. Waterborne nutrient budgets for the
riparian zone of an agricultural watershed.
Agricultural, Ecosystems and Environment
10:371-384.
Lowrance, R.R., R.L., Todd, J. Fail, O.
Hendrickson, R. Leonard, and L.E. Asmussen.
1984. Riparian forests as nutrient filters in
agricultural watersheds. Bioscience 34:374-
377.
Magette, W.L., R.B. Brinsfield, R.E. Palmer, and
J.D. Wood. 1989. Nutrient and sediment
removal by vegetated filter strips. Trans
ASAE 32:663-667
McElroy, A.D., S.Y. Chiu, J.W. Nebgen, A.
Aleti, and F.W. Bennett. 1976. Loading func-
tions for assessment of water pollution from
nonpoint sources. EPA-600/2-76-151.
USEPA, Washington DC.
Meentemeyer, V. 1989. Geographical perspec-
tives of space, time, and scale. Landscape
Ecology 3(3/4):163-173.
Milne, A.K. 1988. Change direction analysis
using Landsat imagery: a review of method-
ology. Proceedings of the IGARSS
Symposium, Edinburgh, Scotland, September
13-16, pages 541-544, (IEEE 88CH2479-6)
Milne, B.T. 1992. Spatial aggregation and neu-
tral models hi fractal landscapes. Amer. Nat.^
139: 32-57.
Milne, B.T. 1993. Analysis for Landscape
Evaluation and Characterization. Pp. 127-
140, in Jensen, M.E., and P.S. Bourgeron
(eds.), Eastside Forest Ecosystem Health
Assessment, Vol n, Ecosystem Management:
Principles and Applications. U.S.D.A. Forest
Service, Missoula, Montana.
Minshall, G.W. 1993. Stream-Riparian
Ecosystems: Rationale and Methods for
Basin-Level Assessments of Management
Effects. Pp. 149-174, in Jensen, M.E., and
P.S. Bourgeron (eds.), Eastside Forest
Ecosystem Health Assessment, Vol II,
50
-------
Ecosystem Management: Principles and
Applications. U.S.D.A. Forest Service,
Missoula, Montana.
Mwalyosi, R.B.B. 1991. Ecological evaluation
for wildlife corridors and buffer zones for
Lake Manyara National Park, Tanzania, and
its immediate environment. Biol. Conserv.
57:171-186.
Mouat, D.A., G.G. Mahin, and J.L. Lancaster.
1993. Remote sensing techniques in the
analysis of change detection. Geocarta
8(2):39-50.
Naiman, R.J., and H. Decamps, eds. 1990. The
ecology and management of aquatic-terrestri-
al ecotones. The Parthenon Publishing Group,
Park Ridge, NJ.
Naiman, R.J., H. Decamps, and M. Pollock.
1993. The role of Riparian Corridors in
Maintaining Regional Biodiversity.
Ecological Applications Vol. 3, No. 2.
Nelson, R.F. 1983. Detecting forest canopy
change due to insect activity using Landsat
MSS. Photogrammetric Engineering and
Remote Sensing, 49(9): 1303-1314.
Noss, R.F. 1990. Indicators for monitoring bio-
diversity: a hierarchical approach.
Conservation Biology, 4(4):355-364.
Odum, E.P. 1978. Opening address: ecological
importance of the riparian zone. In:
Strategies for protection and management of
floodplain wetlands and other riparian ecosys-
tems. USDA Forest Service General
Technical Report, WO-12.
Office of Technology Assessment (OTA). 1987.
Technologies to maintain biological diversity.
U.S. Government Printing Office,
Washington, DC.Omernick, J.M. 1986.
Ecoregions of the United States. Ann. Assoc.
Amer. Geogr. 77(1).
O'Neill, R.V. 1988. Hierarchy theory and glob-
al change. Pp. 29-45, hi R.T. Rosswall, R.G.
Woodmansee, and P.O. Risser (eds.), Spatial
and Temporal Variability in Biospheric and
Geospheric Processes. John Wiley and Sons,
New York.
O'Neill, R.V. 1989. Perspectives in Hierarchy
and Scale. Pp. 140-156, in J. Roughgarden,
R.M. May, and S.A. Levin (eds.), Perspectives
in Ecological Theory. Princeton Univ. Press,
Princeton, NJ.
O'Neill, R.V., D.L. DeAngelis, J.B. Waide, and
T.F.H. Allen. 1986. A hierarchical concept of
ecosystems. Princeton University Press,
Princeton, NJ.
O'Neill, R.V., J.R. Krummel, R.H. Gardner, G.
Sugihara, B. Jackson, D.L. DeAngelis, B.T.
Milne, M.G. Turner, B. Zygmunt, S.W.
Christensen, V.H. Dale, and R.L. Graham.
1988a. Indices of landscape pattern.
Landscape Ecology 1:153-162.
O'Neill, R.V., B.T. Milne, M.G. Turner, and R.H.
Gardner. 1988b. Resource utilization scales
and landscape pattern. Landscape Ecology 2:
63-69.
O'Neill, R.V., A.R. Johnson, and A.W. King.
1989. A hierarchical framework for the analy-
sis of scale. Landscape Ecology 3:193-205
O'Neill, R.V., R. H. Gardner, B.T. Milne, M.G.
Turner, and B. Jackson, 1991a. Heterogeneity
and spatial hierachies. Pp. 85-96, hi J. Kolasa
and S.T.A. Pickett (eds.) Ecological
Heterogeneity. Springer-Verlag, NY.
O'Neill, R.V., S.J. Turner, VI. Cullinen, D.P.
Coffin, T. Cook, W. Conley, J. Brunt, J.M.
Thomas, M.R. Conley, and J. Gosz. 1991b.
Multiple landscape scales: an intersite com-
parison. Landscape Ecology 5:137-144.
O'Neill, R.V., R.H. Gardner, M.G. Turner, and
W.H. Romme. 1992 Epidemiology theory and
disturbance spread on landscapes. Landscape
Ecology 7:19-26.
O'Neill, R.V., C. Hunsaker, and D. Levine.
1992b. Monitoring challenges and innovative
ideas. Pp. 1443-1460, hi D.H. McKenzie,
D.E. Hyatt, and VJ. McDonald (eds.),
Ecological indicators. Elsevier Sci. Publ. Ltd,
London.
Overton, W.S., D. White, and D.L. Stevens.
1990. Design report for EMAP
(Environmental Monitoring and Assessment
Program) EPA/600/S3-91/051.
Pacala, S.W. 1987. Neighborhood models of
plant population dynamic. UI Models with
spatial heterogeneity in the physical environ-
ment. Theor. Pop. Biol. 31:359-392.
Paine, R.T., and S.A. Levin. 1981. Jutertidal
landscapes: disturbance and the dynamics of
patches. Ecological Monographs 51:145-178.
51
-------
Peterjohn, W.T., andD.L. Correll 1984. Nutrient
dynamics in an agricultural watershed:
Observations on the role of a riparian forest.
Ecology 65:1466-1475.
Phillips, J.D. 1989a. Nonpoint source pollution
control effectiveness of riparian forests along
a coastal plain river. J Hydrology 110:
221-237.
Phillips, J.D. 1989b. An evaluation of the fac-
tors determining the effectiveness of water
quality buffer zones. J Hydrology 107:
133-145.
Quarnby, N.A., and J.L. Cushine. 1985.
Monitoring urban land cover changes at the
urban fringe from SPOT HRV imagery in
southeast England. Inter. J. Remote Sens.
10:953-963.
Regan, R.M., and J.D. FeUows 1980. An
overview of remote sensing based information
systems for hydrologic modeling. Proc 14th
Int. Sym. on Remote Sensing of the
Environment Vol I. Environmental Research
Institute of Michigan, Ann Arbor.
Runkle, J.R. 1985. Disturbance regimes in tem-
perate forests. Pages 17-34 in S.T.A. Pickett
and P.S. White, eds. The ecology of natural
disturbance and patch dynamics. Academic
Press, New York.
Saunders, D.A., RJ. Hobbs, and C.R. Margules.
1991. Biological consequences of ecosystem
fragmentation: a review. Conservation
Biology 5(l):18-32.
Schlesinger, W.H., J.R Reynolds, G.L.
Cunningham, L.F. Huenneke, W.M. Jarrell,
R.A. Virginia, and W.G. Whitford. 1990.
Biological feedbacks in global desertification.
Science 247:1043-1048.
Schlosser, I.J., and J.R. Karr. 1981a. Water
quality in agricultural watersheds: impact of
riparian vegetation during base flow. Water
Resources Bulletin, 17:233-240.
Schlosser, I.J., and J.R. Karr 198 Ib. Riparian
vegetation and channel morphology impact on
spatial patterns of water quality in agricultural
watersheds. Environmental Management
5:233-243.
Schlosser, IJ. 1991. Stream fish ecology: a
landscape perspective. BioScience 41:704-
711.
Shanholtz, V.O., E.R. Yaghow, J.M. Flagg, R.K.
Byler, S. Mostaghimi, and T.A. Dilliha. 1988.
Agricultural pollution potential database for
Culpepper Soil and water conservation dis-
trict. Interim report VirGIS 88-2. Dept of
Conservation and Historic Resources,
Richmond, VA. 71pp.
Shepard, J.R. 1964. A concept of change detec-
tion. Photogrammetric Engineering, 30(4):
648-651.
Shlisky, A.J. 1993. Multiscale Ecosystem
Analysis and Design hi the Pacific Northwest
Region: The Umatilla National Forest
Restoration Project. Pp. 259-268, in Jensen,
M.E., and P.S. Bourgeron (eds.), Eastside
Forest Ecosystem Health Assessment, Vol IT,
Ecosystem Management: Principles and
Applications. U.S.D.A. Forest Service,
Missoula, Montana-Shmida, A., and S. Ellner.
1984. Coexistence of species with similar
niches. Vegetation 58:29-55.
Singh, A. 1989. Digital change detection tech-
niques using remotely-sensed data.
International Journal of Remote Sensing
10(6):989-1003.
Soule, M.E., A.C. Alberts, and D.T. Bolger.
1992. The effects of habitat fragmentation on
chaparral plants and vertebrates. Oikos 63:39-
47.Sprugel, D.G., and F.H. Bormann 1981.
Natural disturbance and the steady state in
high-altitude balsam fir forests. Science,
211:390-393.
Stauffer, M.L., and R.L. McKinney. 1978.
Landsat image differencing as an automated
change detection technique. Report contract
NAS-24350, General Electric Company,
Beltsville, Maryland.
Stauffer, D. 1985. Introduction To Percolation
Theory. Taylor and Francis, London.
Stoms, D.M., and J.E. Estes. 1993. A remote
sensing research agenda for mapping and
monitoring biodiversity. International Journal
of Remote Sensing, 14(10):1839-1860.
Terborgh, J. 1992. Why American songbirds are
vanishing. Sci. Am. (May):98-104.
Thomas, J.W., C. Maser, and J.E. Rodiek. 1979.
Wildlife habitat in managed rangelands - the
Great Basin of Southeastern Oregon. Riparian
zones. USDA Forest Service General
Technical Report PNW-80, Portland, OR.
52
-------
Tucker, C.J. 1979. Red and photographic
infrared linear combinations for monitoring
vegetation. Remote Sensing of Environment,
8:127-150.
Turner, M.G. 1987a. Spatial simulations of
landscape change in Georgia: a comparison of
3 transition models. Landscape Ecology 1:29-
36.
Turner, M.G. (ed.) 1987b. Landscape hetero-
geneity and disturbance. Springer-Verlag,
NY.
Turner, M.G. 1989. Landscape ecology: the
effect of pattern on process. Annual Review
of Ecology and Systematics 20:171-197.
Turner, M.G., and C.L. Ruscher. 1988. Changes
in landscape patterns in Georgia, USA.
Landscape Ecol. 1:241-251.
Turner, S.J., R.V. O'Neill, W. Conley, M.R.
Conley, and H.C. Humphries. 1991. Pattern
and Scale: Statistics for Landscape Ecology.
Pp. 17-41, in M.G. Turner and R.H. Gardner
(eds.), Quantitative Methods in Landscape
Ecology. Springer-Verlag, NY.
Ulfstrand, S. 1992. Biodiversity - how to reduce
its decline. Oikos 63:3-4.
United States Department of Agriculture
(USDA). 1951. Soil Survey Manual. U.S.
Department of Agriculture, Agricultural
Handbook #18. Washington, DC.
Urban, D.L., R.V. O'Neill, and H.H. Shugart, Jr.
1987. Landscape Ecology. BioScience
37:119-127.
Van Der Zee, F.F., J. Wiertz, CJ.F. Ter Braak,
R.C. Van Apeldoorn, and J. Vmk. 1992.
Landscape change as a possible cause of the
badger Meles meles L. decline in the
Netherlands. Biol. Conserv. 61:17-22.
Weismiller, R.A., SJ. Kristof, O.K. Scholz, P.E.
Anuta, and S.A. Momin. 1977. Change
detection in coastal zone environments.
Photogrammetric Engineering and Remote
Sensing, 43(12): 1533-1539.
White, P.S. 1979. Pattern, process, and natural
disturbance in vegetation. Botanical Review
45:229-299.
Wickham, J.D. 1988. Land cover change map-
ping using Landsat Thematic Mapper data.
Report contract no. DACA76-86-C-0018,
U.S. Army Engineer Topographic
Laboratories, ETL-GS LC, Fort Belvoir,
Virginia.
WickhamJ.D.,andD.J.Norton. 1994. Mapping
and analyzing landscape patterns. Landscape
Ecology 9 (1): 7-23
Whitmore, R.C., and G.G. Ice. 1984. A review
of models for estimating the effect of forestry
management practices on yield and transport
of materials to receiving waters. National
Council of the Paper Industry for Air and
Stream Improvement. Technical Bulletin 420,
New York. 153pp.
Whittaker, R.H. 1975. Communities and
Ecosystems. 2nd Ed. MacMillan and
Company, New York.Wiens, J.A. 1985.
^Vertebrate responses to environmental patchi-
^Sfless in arid and semiarid ecosystems. In:
Pickett, S.T.A., and P.S. White, eds. The
Ecology of Natural Disturbances and Patch
Dynamics. Academic Press, New York:
169-193.
Wiens, J.A., and B.T. Milne. 1989. Scaling of
'landscapes' hi landscape ecology, or land-
scape ecology from a beetle's perspective.
Landscape Ecology 3: 87-96.
Wilson, E.O. (ed.) 1988. Biodiversity. National
Academy Press, Washington, D.C.
53
•frll.S. GOVERNMENT PRINTING OFFICE: 1994 - 550401/80393
-------
-------
-------
------- |