Content uploaded by Reed Noss
Author content
All content in this area was uploaded by Reed Noss on Mar 10, 2020
Content may be subject to copyright.
895
Conservation Biology, Pages 895–908
Volume 16, No. 4, August 2002
Contributed Papers
A Multicriteria Assessment of the Irreplaceability
and Vulnerability of Sites in the Greater
Yellowstone Ecosystem
REED F. NOSS,* CARLOS CARROLL, KEN VANCE-BORLAND,
AND GEORGE WUERTHNER
Conservation Science, Inc., 7310 NW Acorn Ridge, Corvallis, OR 97330, U.S.A.
Abstract:
We conducted a systematic conservation assessment of the 10.8-million-ha Greater Yellowstone Eco-
system (GYE), integrating three basic approaches to conservation planning: protecting special elements, rep-
resenting environmental variation, and securing habitat for focal species (grizzly bear [
Ursus arctos
], wolf
[
Canis lupus
], and wolverine [
Gulo gulo
]). Existing protected areas encompass 27% of the GYE but fail to cap-
ture many biological hotspots of the region or to represent all natural communities. Using a simulated an-
nealing site-selection algorithm, combined with biological and environmental data based on a geographic in-
formation system and static (habitat suitability) and dynamic ( population viability) modeling of focal
species, we identified unprotected sites within the GYE that are biologically irreplaceable and vulnerable to
degradation. Irreplaceability scores were assigned to 43 megasites (aggregations of planning units) on the
basis of nine criteria corresponding to quantitative conservation goals. Expert opinion supplemented quanti-
tative data in determining vulnerability scores. If all megasites were protected, the reserved area of the GYE
would expand by 43% (to 70%) and increase protection of known occurrences of highly imperiled species by
71% (to 100%) and of all special elements by 62% (to 92%). These new reserves would also significantly in-
crease representation of environmental variation and capture critical areas for focal species. The greatest
gains would be achieved by protecting megasites scoring highest in irreplaceability and vulnerability. Protec-
tion of 15 high-priority megasites would expand reserved area by 22% and increase the overall achievement
of goals by 30%. Protection of highly imperiled species and representation of geoclimatic classes would in-
crease by 46% and 49%, respectively. Although conservation action must be somewhat opportunistic, our
method aids decision-making by identifying areas that will contribute the most to explicit conservation goals.
Evaluación de Criterios Múltiples de la Irremplazabilidad y Vulnerabilidad de Sitios en el Ecosistema Mayor de
Yellowstone
Resumen:
Realizamos una evaluación sistemática de conservación de las 10.8 millones de Ha del Eco-
sistema Mayor de Yellowstone (GYE), empleando las tres rutas de protección de los elementos especiales, la
representación de la variación ambiental y asegurando hábitat para especies focales (el oso grizzli [
Ursus
arctos
], el lobo [
Canis lupus
], y el carcayú glotón [
Gulo gulo
]). Las áreas protegidas abarcan 27% del GYE pero
no logran capturar muchas áreas prioritarias para la conservación de la región ni representar a todas las co-
munidades naturales. Usando un algoritmo simulado de selección de sitio, combinado con datos biológicos y
ambientales basados en GIS y empleando el modelado estático (aptitud del hábitat) y dinámico (viabilidad
poblacional) de especies focales, identificamos sitios desprotegidos dentro del GYE que son biológicamente ir-
reemplazabilidad y vulnerables a la degradación. Se asignaron puntajes de irreemplazabilidad a 43 megasi-
tios (agregaciones de unidades de planeación) en base a nueve criterios correspondientes a las metas cuanti-
tativas de conservación. Si todos los megasitios fueran protegidos, el área reserva del GYE se expandiría en
un 43% (a 70%) e incrementaría la protección de ocurrencias conocidas de especies altamente en peligro en
un 62% (a 92%). Estas nuevas reservas también incrementarían significativamente la representación de la
*
Current address: University of Central Florida, Department of Biology, 4000 Central Florida Blvd., Orlando, FL 32816-2368, email reed noss@
conservationscience.com
Paper submitted August 20, 2001; revised manuscript accepted January 30, 2002.
896
Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
Introduction
Systematic planning on a regional scale has become the
standard approach of conservation organizations and agen-
cies worldwide (Noss 1983; Pressey 1994; Dinerstein et al.
1995; Noss et al. 1997; Groves et al. 2000; Pressey &
Taffs 2001). Systematic conservation planning is superior
in many ways to opportunistic or politically biased ap-
proaches, which have resulted in a skewed distribution of
protected areas (Pressey et al. 1993; Scott et al. 2001).
Among the key attributes of systematic conservation plan-
ning are explicit goals, quantitative targets, explicit meth-
ods for locating new reserves to complement existing ones,
and rigorous criteria for implementing conservation ac-
tion (Margules & Pressey 2000).
The Greater Yellowstone Ecosystem (GYE) was first
defined as the area necessary to sustain the Yellowstone
population of grizzly bears (
Ursus arctos
; Craighead
1979). Today, the GYE is the southernmost area in
North America that still contains a full suite of native
carnivores, along with other wilderness qualities (Clark
et al. 1999). Although the GYE is not a biological hot-
spot on a global or continental scale (Ricketts et al. 1999;
Myers et al. 2000), it presents an opportunity that most
hotspots do not—to conserve a reasonably intact tem-
perate ecosystem. Nevertheless, biodiversity in the re-
gion is threatened. The GYE’s scenic qualities have at-
tracted a burgeoning human population, the impacts of
which now rival the traditional threats of resource ex-
traction. Population-growth rates for the 20 counties
within the GYE averaged 14% between 1990 and 1999
and ranged up to 66% (Greater Yellowstone Coalition,
unpublished data). The GYE is unique in that large core
refugia lie in close proximity to a rapidly growing hu-
man population.
We conducted a conservation assessment of the GYE to
serve four well-accepted goals of conservation (Noss &
Cooperrider 1994): (1) to represent ecosystems across their
natural range of variation; (2) to maintain viable popula-
tions of native species; (3) to sustain ecological and
evolutionary processes; and (4) to build a conservation
network that is resilient to environmental change. In pur-
suit of these goals we integrated three basic approaches
to conservation planning: (1) protection of special ele-
ments, including imperiled species and communities; (2)
representation of a broad spectrum of environmental varia-
tion, including that of vegetation, geoclimatic, and aquatic
classes; and (3) protection of critical habitats of focal spe-
cies (Lambeck 1997; Miller et al. 1998–1999), whose needs
help planners address issues of habitat area, configura-
tion, and quality. Together, these three approaches con-
stitute a comprehensive approach to conservation plan-
ning (Noss et al. 1999).
Using a simulated annealing site-selection algorithm ap-
plied to several classes of conservation targets, we identi-
fied sites within the GYE that have the most to lose if not
protected. To identify priority sites, we relied on two key
concepts: irreplaceability and vulnerability (Pressey et al.
1994; Margules & Pressey 2000; Pressey & Cowling 2001).
Irreplaceability provides a quantitative measure of the rela-
tive contribution made by different areas to reaching con-
servation goals, thus helping planners choose among al-
ternative sites. We assessed vulnerability on the basis of
expert opinion and consensus about the threats faced by
each site, taking into account available quantitative data.
Our findings provide a template for determining conser-
vation priorities and making land-allocation decisions
with awareness of the trade-offs involved.
Study Area
We defined the study area (Fig. 1) in consultation with
the Greater Yellowstone Coalition, with consideration of
mountain ranges, watersheds, wildlife migration routes, and
other features. The core of this 10.8-million-ha region is
the 890,000-ha Yellowstone National Park (YNP), the
134,000-ha Teton National Park and John D. Rockefeller
Memorial Parkway, and an additional 1.6 million ha of
federally designated wilderness. Altogether, 36% of the GYE
is private land and 64% is public land. Protected areas
recognized by the U.S. Geological Survey (USGS) Gap
Analysis Program (GAP) (Scott et al. 1993, 1996) consti-
tute 2.9 million ha, 27% of the GYE.
Lower elevations of the GYE are generally treeless ex-
cept along streams and are dominated by grass-shrub com-
munities. Coniferous forests dominate middle to upper
elevations. Douglas-fir (
Pseudotsuga menziesii
) is the
dominant low-elevation tree, whereas Engelmann
spruce (
Picea engelmannii
), subalpine fir (
Abies lasio-
carpa
), and lodgepole pine (
P. contorta
) compose mid-
elevation forests. Spruce-fir forest would dominate more
variación ambiental y capturarían áreas críticas para especies focales. Las mayores ganancias se obtendrían
mediante la protección de megasitios con los puntajes más altos de irreemplazabilidad y vulnerabilidad. La
protección de 15 megasitios altamente prioritarios expandiría el área de reserva en un 22% e incrementaría
el éxito general de las metas en un 30%. La protección de especies altamente en peligro y la representación de
clases geoclimáticas podría incrementar en un 46% y un 49% respectivamente. A pesar de que la acción de
conservación debe ser de alguna manera oportunista, nuestro método ayuda a la toma de decisiones al iden-
tificar áreas que contribuirán más a explicar las metas de conservación.
Conservation Biology
Volume 16, No. 4, August 2002
Noss et al. Irreplaceability and Vulnerability of Greater Yellowstone Sites
897
of the area were it not for stand-replacement fires that
favor lodgepole pine. Limber pine (
P. flexilis
) occurs
throughout on dry, windy sites, and aspen (
Populus
tremuloides
) and willow (
Salix
spp.) occur in wetter
areas. At high elevations, whitebark pine (
P. albicaulis
)
is common. Extensive tracts of alpine tundra occur
above timberline (Knight 1994).
Methods
Planning Units
The building blocks of a conservation plan are the sites
that are compared to one another. We used sixth-level
catchments (hydrologic units; USGS 2001) as planning
units because they are ecologically relevant and at a con-
venient scale for regional planning. Nevertheless, sixth-
level catchments had not been delineated for most of
the GYE. We used the BASINS function in ArcInfo GRID
geographic information system (GIS) software, which is
based on a 90-m digital-elevation model to create pseudo
(modeled) sixth-level catchments. To better conform the
resulting polygons to recognized catchments, we
merged them with USGS fifth-level catchments. We elim-
inated polygons smaller than 2000 ha and divided large
polygons to avoid potential species-area effects.
To distinguish existing protected areas, we merged the
catchment polygons with USGS GAP polygons of man-
agement status 1 (strictly protected) and 2 (moderately
protected). This procedure resulted in 1908 planning
units (range: 13–43,564 ha;
5,692 ha). (Smaller units
x
Figure 1. The Greater Yellowstone
Ecosystem, showing protected areas
and the expanded study area for
focal species.
898
Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
were catchments located partly within existing protected
areas. Only portions of the catchments that fell outside
protected areas were considered planning units.) After
applying a site-selection algorithm (see below), we ag-
gregated selected units into “megasites” for priority set-
ting. These larger sites are likely to be more effective
conservation units than sixth-level watersheds. Bound-
aries of fourth-level watersheds and other natural fea-
tures were used to define megasite boundaries.
The SITES Selection Algorithm
We used SITES (version 1.0; Andelman et al. 1999) to as-
semble and compare alternative portfolios of planning
units. SITES can use either of two algorithms—the greedy
heuristic or simulated annealing with iterative improve-
ment—to assemble portfolios (Possingham et al. 2000).
We used simulated annealing, the more efficient of the
two. This algorithm does not guarantee an optimal solu-
tion, which is prohibitive in computer time for large,
complex data sets. Rather, the algorithm attempts to
minimize portfolio “cost” while maximizing attainment
of conservation goals: cost
area
species penalty
boundary length, where area encompasses all planning
units selected for the portfolio, species penalty is im-
posed for failing to meet target goals, and boundary
length includes the total boundary of the portfolio (An-
delman et al. 1999; Possingham et al. 2000). The bound-
ary-length modifier is a notable improvement on early
site-selection algorithms, which often neglected the con-
figuration of sites and resulted in fragmented portfolios
that are difficult to manage (Briers 2002; McDonnell et al.
2002). SITES minimizes portfolio cost by selecting the
smallest overall area needed to meet target goals and by
selecting planning units that are clustered or adjacent to
existing reserves rather than dispersed.
We had SITES perform 1,000,000 iterative attempts to
find the minimum cost solution per simulated annealing
run and perform 10 such runs for each of dozens of al-
ternative scenarios. Alternative scenarios were con-
structed by varying target goals and inputs to the cost
function. Often, several different portfolios will meet
goals almost equally well, providing flexibility to the
planning process (Leslie et al. 2002). Varying goals and
inputs, and examining alternative portfolios, allow ex-
pert knowledge to be incorporated into an interactive
decision-making process. The final set of goals and the fi-
nal portfolio were ones we and the client (the Greater
Yellowstone Coalition) felt comfortable with.
Special Elements
Following guidelines established by The Nature Conser-
vancy (Groves et al. 2000; Stein & Davis 2000), we as-
sembled element-occurrence data for the GYE from
natural heritage programs in Montana, Idaho, and Wyo-
ming. After we excluded occurrences of species or com-
munities last observed prior to 1982 or ranked as nonvi-
able or nonbreeding by the heritage programs, 2303
occurrences of 435 species and communities remained
(Fig. 2), 203 of them for the 55 species and communi-
ties with conservation status ranks of G1 (critically im-
periled globally) or G2 (imperiled globally). Considering
the disparate area requirements and databases describ-
ing the distributions of different taxa, we divided the oc-
currence data into four groups for separate SITES analy-
sis: local-scale species, vulnerable and declining bird
species, coarse- and regional-scale aquatic (fish) species,
and plant communities (Groves et al. 2000; Poiani et al.
2000; Noss et al. 2001). We set goals for capturing 100%
of the G1 and G2 occurrences in all groups and at least
50% of occurrences of less-threatened elements.
We made 10 runs for each target group, using the
SITES “sum runs” option (Andelman et al. 1999). Output
from sum runs included the number of times each plan-
ning unit was included in the 10 portfolios, and the
“best” (lowest-cost) portfolio of the 10. Contiguous or
near-contiguous planning units with sum-runs values of
1 were aggregated into megasites. The number of times
the planning units in a megasite were selected during
the 10 runs was used, in part, to determine irreplaceabil-
ity (similar to the method of Leslie et al. [2002]).
Representation
We used a combination of vegetation types derived from
satellite imagery and mapped by the state GAP programs
and a new classification of physical (abiotic, geoclimatic)
classes to represent variation in terrestrial ecosystems. A
combination of biotic and abiotic surrogates is likely to
capture more variation than either class alone (Kirk-
patrick & Brown 1994), especially because each GAP
vegetation type encompasses considerable internal het-
erogeneity. Moreover, representing a spectrum of physi-
cal substrates and associated vegetation—ideally along
intact gradients—may facilitate shifts in species distribu-
tions in response to climate change (Noss 2001).
We merged state-level GAP vegetation maps into a sin-
gle map portraying 39 vegetation types in the GYE. We
set goals for capturing at least 25% of the area of each
wetland vegetation type—lowland riparian, mountain ri-
parian, water, wetland, wet meadow—and 15% of all oth-
ers, with the justification that wetlands are considered of
higher biological value in the region (Patten 1998). We
used the major components of climate variation in the re-
gion to classify geoclimatic types in ArcInfo GIS: (1) mean
annual precipitation, (2) spring precipitation, (3) mean
annual low temperature, (4) mean annual high tempera-
ture, and (5) the difference between winter mean low
temperature and summer mean high temperature (Daly
et al. 1994), in addition to mean annual growing degree
days. Edaphic variables, assumed to be of secondary im-
Conservation Biology
Volume 16, No. 4, August 2002
Noss et al. Irreplaceability and Vulnerability of Greater Yellowstone Sites
899
portance in influencing species distributions, were soil
depth, water-holding capacity, and organic carbon con-
tent, derived from the STATSGO soils database (Soil Sur-
vey Staff 1992). A cluster analysis of the nine climate and
soil variables recognized 38 geoclimatic classes. We set
goals for capturing at least 15% of the area of each type.
For aquatic representation, we used the aquatic classi-
fication developed by The Nature Conservancy (M. Lam-
mert, personal communication). We applied two levels of
classification: (1) aquatic macrohabitats, identified at the
stream-reach level, and (2) aquatic ecological systems,
identified at the watershed to basin level. Both classifica-
tions utilize four components: stream size (headwater to
large river); elevation (low to alpine); stream gradient
(low to very steep); and dominant geology (coarse, po-
rous, nonporous). Aquatic macrohabitats were classified
by specific portions of the range of each component
(e.g., “very steep alpine headwater in coarse geology”).
Aquatic ecological systems, being aggregations of mac-
rohabitats, represent a greater range of gradients. We
stratified ecological systems by the macrohabitats they
contain and set goals for representing at least 20% of
each of the 654 strata.
Focal Species
Our study placed greater emphasis on species viability
than most previous multicriteria conservation plans. Al-
Figure 2. Element occur-
rences (imperiled species and
communities) from state nat-
ural heritage programs. Black
dots represent occurrences of
elements critically imperiled
globally (G1) and imperiled
globally (G2), and gray dots
represent less imperiled ele-
ments.
900
Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
though a comprehensive set of focal species would en-
compass taxa sensitive to a range of environmental fac-
tors (Lambeck 1997; Miller et al. 1998–1999), we
selected four area-limited carnivores and an ungulate:
grizzly bear, gray wolf (
Canis lupus
), wolverine (
Gulo
gulo
), lynx (
Felis lynx
), and elk (
Cervus elaphus
). Re-
sults for the first three of these species, for which the
models are most robust, are reported here (for the full
analysis see Noss et al. 2001). Because of the large spa-
tial scales over which metapopulation dynamics of these
species operate, we analyzed an expanded study region
of 31.6 million ha (Fig. 1).
We used GIS data on distribution and habitat charac-
teristics to construct static habitat-suitability models (i.e.,
resource selection functions; Boyce & McDonald 1999)
for each focal species, with methods developed by Car-
roll et al. (2001). These results were compared with
those from dynamic models that considered regional
population dynamics within a multiregional context
(Noss et al. 2001). Species-distribution data included
sightings records of wolverines, radiotelemetry locations
of grizzly bears, and the boundaries of wolf-pack territo-
ries. Habitat data included vegetation, satellite-imagery
metrics, topography, climate, and variables related to
human impacts (e.g., road density; Mladenoff et al. 1995;
Merrill et al. 1999). We used multiple logistic regression
to compare habitat variables at telemetry or sighting lo-
cations with those at random points. We used the coeffi-
cients from the final model to calculate a resource selec-
tion function (RSF) for used (occurrences) and available
(random) resources (Manly et al. 1993; Boyce & McDonald
1999).
We performed population viability analyses with the
program PATCH (Schumaker 1998). PATCH links the
survival and fecundity of individual animals to GIS vari-
ables corresponding to mortality risk and habitat pro-
ductivity, measured within individual or pack territories.
The model tracks the population as individuals are born,
disperse, and die and allows the landscape to change
through time. Hence, the user can predict the conse-
quences of landscape change for population viability and
identify probable sources and sinks. Our landscape-
change scenarios used estimates of potential change in
human-associated impact factors (e.g., roads and human
population) during the period 2000–2025, given in-
creased development on either private and public lands
or on private lands only.
We set a SITES goal of habitat sufficient to support
75% of the current potential population of each species,
as defined by the RSF. We then compared SITES solu-
tions with results from the PATCH model. We scored
areas selected by SITES as to their irreplaceability and
vulnerability by overlaying megasite boundaries on the
PATCH results. Irreplaceability in this context is the
value of an area as source habitat, defined by population
growth rate (lambda). Vulnerability is the predicted de-
cline in lambda over the next 25 years. Dynamic model
results also contributed one of nine criteria determining
overall megasite irreplaceability. PATCH irreplaceability
in the composite scores was an average of lambda values
for the focal species, weighted by the likelihood that a
site was occupied by the species. We also ran the
PATCH model on potential future landscapes that in-
cluded increased protection and restoration of mega-
sites.
Expert Assessment
Quantitative data by which to evaluate conservation op-
tions are always limited. We supplemented quantitative
measures of conservation value with expert opinion, ap-
plying a combined approach of one-on-one interviews
followed by a workshop. G. Wuerthner interviewed 124
experts on the ecology and conservation issues of the
GYE during 1999–2000. Interviews included discussion
of habitat conditions, imperiled species, threats, moni-
toring, survey, and management. After producing our
draft report, we participated in a workshop to present
our results, evaluate alternative portfolios, and assign
vulnerability scores to selected megasites. The workshop
was attended by experts on conservation issues across
the GYE.
Megasite Ranking
Many potential methods exist for estimating irreplace-
ability, which cannot be measured directly (Ferrier et al.
2000; Pressey & Taffs 2001; Leslie et al. 2002). Because
our assessment considered multiple values of megasites
and attempted to achieve a broad set of conservation
goals, we assigned irreplaceability values to megasites
based on nine criteria assessed as contributions to the
following goals (each considered a minimum threshold):
(1) protects 50% (or 100% for G1/G2) of viable occur-
rences of imperiled local-scale species; (2) protects 50%
(or 100% for G1/G2) of viable occurrences of vulnerable
and declining bird species; (3) protects 50% (or 100%
for G1/G2) of viable occurrences of coarse-scale and re-
gional-scale aquatic (fish) species; (4) protects 50% (or
100% for G1/G2) of viable occurrences of plant commu-
nities; (5) represents 25% of the area of each wetland
vegetation type and 15% of the area of each other vege-
tation type in the region; (6) represents 15% of the area
of each geoclimatic class in the region; (7 ) represents
20% of the length of each aquatic (stream) habitat type
in the region; (8) protects habitat capable of supporting
75% of the population of each focal species that cur-
rently could be supported in the region, as identified by
RSF models; and (9) maintains viable populations of fo-
cal species over time, as determined by PATCH. Each score
Conservation Biology
Volume 16, No. 4, August 2002
Noss et al. Irreplaceability and Vulnerability of Greater Yellowstone Sites
901
is an average of the predicted lambda values, weighted
by the likelihood that the site was occupied by the spe-
cies.
Each megasite was scored from 0 to 10 for each of the
nine criteria. For criteria 1–8, the number of times (out
of 10) that individual planning units were selected in
SITES sum runs was used to calculate an area-weighted
mean score for each megasite. For criterion 9, entire me-
gasites were scored as units. A total irreplaceability
score was calculated for each megasite by summing the
area-weighted mean scores for the nine criteria. These
total scores were then rescaled to range from approxi-
mately 1 to 100. We weighted the nine criteria equally in
this exercise, although planners could apply different
weightings to emphasize particular criteria.
Vulnerability cannot be estimated as objectively as ir-
replaceability (Pressey et al. 2000). We prepared a site
description for each megasite (Noss et al. 2001), which
included a discussion of known threats and management
issues, as derived from interviews with the 124 experts.
We also acquired region-wide data on such threat surro-
gates as human population and development trends
(e.g., Theobald 2001). We then assigned a preliminary
vulnerability score from 1 to 100 to each megasite based
on multiple criteria, including the proportion of the site
in private versus public ownership; presence of active
grazing, mining, oil and gas, or timber leases or potential
for such in the near future; road density and trends; hu-
man population and housing density and trends; disrup-
tive recreational uses and trends, among others. Prelimi-
nary vulnerability scores were revised by consensus by
participants in the workshop, and those revised scores
were rescaled to range approximately from 1 to 100.
Following Margules and Pressey (2000), we plotted
megasites on a graph of irreplaceability (y-axis) versus
vulnerability (x-axis) and divided the graph into four
quadrants. The upper right quadrant, which encom-
passes megasites with high irreplaceability and high vul-
nerability, is the highest priority for conservation. This
top tier of megasites is followed by the upper left and
lower right quadrants (moderate priority), and finally by
the lower left quadrant, encompassing megasites that
are relatively replaceable and face less severe threats.
Results and Discussion
Our assessment resulted in a portfolio of planning units
grouped into 43 megasites, which collectively met or ex-
ceeded the stated conservation goals for the three ap-
proaches of special elements, representation, and focal
species. These megasites constitute areas of public and
private land that currently lack formal protection. Me-
gasites range in size from 11,331 to 315,662 ha (
109,268 ha) and total 4,573,047 ha (42% of the GYE).
Private lands constitute 36% of the total portfolio area. If
x
combined with existing protected areas (totaling 2,889,518
ha), our portfolio would bring the total protected area to
7,462,565 ha, nearly 70% of the GYE.
Special Elements
The complete portfolio, together with existing protected
areas, captured 100% of the 236 documented occurrences
of G1 and G2 species and communities in the GYE, as tar-
geted, compared with 29% in existing protected areas. On
average, protection of element occurrences for the four
classes of special elements—local-scale species, birds,
fish, and plant communities—increased by 62% with this
portfolio.
The proposed portfolio captured occurrences of 80 lo-
cal-scale species that are not recorded in existing pro-
tected areas, including five G1 species (for details see
Noss et al. 2001). The 318 occurrences of vulnerable bird
species in the portfolio, combined with 139 in existing
protected areas, encompassed 457 (86%) of the 534 oc-
currences in the GYE. The 55 occurrences of imperiled
fish species or subspecies in the proposed portfolio,
when added to 24 occurrences in existing protected ar-
eas, encompassed 88% of the 90 known occurrences.
Seven fish species or subspecies not recorded in existing
protected areas were captured by the proposed portfo-
lio. The 341 plant-community occurrences in the pro-
posed portfolio and 148 in existing protected areas cap-
tured 95% of the 515 occurrences in the GYE. Seven G1
plant communities in the proposed portfolio are not
known from existing protected areas.
Representation
Of the three approaches, our proposed portfolio most
fully met stated goals for representation. Current pro-
tected areas met our 15–25% representation goals for
over 61% of the 39 GAP vegetation types and our 15%
goal for 41% of the 39 physical habitat types. Our pro-
posed portfolio met stated goals for 100% of these
classes. Current protected areas met our 20% representa-
tion goal for over 44% of the 654 aquatic habitat strata.
Our proposed portfolio raised this to 98%.
Focal Species
Habitat-suitability modeling produced resource selection
functions for the focal species that predicted association
of individual animals with habitat variables on a regional
scale (Table 1). Of particular management significance
is the negative association of grizzly bear and wolf loca-
tions with road and trail density and the positive associa-
tion of all three species with parks and wilderness areas,
especially compared with private lands. These relation-
902
Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
ships reflect the need of these species for security from
human persecution and harassment (Noss et al. 1996;
Carroll et al. 2001). The wolf model was generally simi-
lar to that for the grizzly bear but differed in the strong
negative association with slopes above 20 degrees and
the lower contrast between parks and wilderness and
private lands, reflecting somewhat greater tolerance of
wolves for human-modified landscapes. The wolverine
model had poorer predictive power than those of the
grizzly bear and wolf, consistent with the scarcity of
field knowledge of habitat relations for this species.
Our dynamic model (PATCH) showed strong source
habitat for grizzly bears in the areas that encompass most
of the current distribution, centered on protected areas in
the core GYE (Fig. 3a). The model suggests that, given
time, bears could expand into adjacent unprotected pub-
lic lands, especially on the eastern and southern periph-
ery of the core. Expansion is limited, however, by the sur-
rounding sink habitat. Assuming that road-building
continues on private and public land over the next 25
years, core areas of the GYE are predicted to remain
strong sources but would no longer be able to support
the peripheral distribution. Essentially, the ring of sinks
that envelops the core GYE becomes increasingly con-
stricted (Fig. 3b). Although we predict a relatively high
probability of persistence for the grizzly bear over the
short to medium term, future landscape change may
greatly reduce the distribution and size of the population.
The dynamic-model results for the wolf were similar
to those for the grizzly bear. Because the wolf can in-
habit semideveloped landscapes, it is more affected by
simulated future development outside the core GYE
(Noss et al. 2001). Currently, potential source areas in-
clude most public lands in the central GYE and some ad-
jacent private lands. Connectivity is reasonably good be-
tween the GYE and areas to the west and north (e.g.,
central Idaho, where another reintroduced wolf popula-
tion exists). Future landscape change threatens to iso-
late wolves in the GYE from adjacent regions and alien-
ate much of the productive lower-elevation habitat that
currently could support wolves. For the wolverine, we
predict a pattern of source and sink habitat similar to
that of the grizzly bear, with a pronounced ring of sinks
surrounding the source habitat of the core GYE (Noss et
al. 2001).
Evaluation of SITES Portfolios Based on the PATCH Model
Our irreplaceability and vulnerability scoring based on
overlaying megasite boundaries on results of the PATCH
models showed broad similarities with the general me-
gasite scoring (Noss et al. 2001). Nevertheless, consis-
tent differences emerged because the PATCH model
considered a broader region (Fig. 1) and accounted for
the landscape context and source-sink dynamics of sites.
For example, sites that emerged as most vulnerable for
carnivores were not necessarily those with the highest
levels of site-specific threat but rather were those whose
continued degradation would pose the greatest risk to
the viability of nearby large source areas that sustain re-
gional populations. Hence, improving conditions in
strong and worsening sinks is potentially as important to
regional population viability as protecting strong
sources. Simulations that included megasites within the
reserve network showed that sites whose protection
would have the greatest effect on the distributions of
focal species were those with high demographic irre-
placeability and vulnerability. Protection of some
smaller megasites would increase occupancy far beyond
their immediate boundaries, particularly for the wolf,
because of its better ability to inhabit semideveloped
landscapes.
Megasite Prioritization
Our procedure for calculating the irreplaceability of me-
gasites necessarily hides information specific to individ-
ual planning units. This information is not lost, however,
and can be accessed readily by planners engaged in site-
level planning and management, a phase beyond the re-
gional scope of this paper. Megasite irreplaceability
scores ranged from 0.3 to 99.5 (
54.9) and vulnera-
bility scores from 1.5 to 98.5 (
50.3). Our analysis
x
x
Table 1. Focal species resource-selection function models for
grizzly bears, wolves, and wolverines of the Greater Yellowstone
Ecosystem.
a
Variable
Grizzly
bear Wolf Wolverine
July brightness
July greenness
July wetness
November brightness
November greenness
November wetness
Annual precipitation cx
b
Annual snowfall cx
Elevation cx
Slope
cx
Elk winter range
Road and trail density
General public land
Wilderness
Park
Road density
public land
Road/trail density
wilderness
Road/trail density
park
November brightness
wetness
a
Selected models are those that explained the most variation in oc-
currences ( locations). Models were highly significant (
p
0.001)
for each species. Variables are shown as positively (
) or negatively
(
) associated with occurrences. See Noss et al. (2001) for model co-
efficients and other details.
b
cx, quadratic, convex up.
Conservation Biology
Volume 16, No. 4, August 2002
Noss et al. Irreplaceability and Vulnerability of Greater Yellowstone Sites
903
resulted in 15 megasites totaling 2.4 million ha in the
high irreplaceability–high vulnerability quadrant 1, giving
them the highest priority for conservation action (Figs.
4 & 5). Ten megasites in high irreplaceability–low vul-
nerability quadrant 2 cover 1.1 million ha; 5 megasites in
low irreplaceability–high vulnerability quadrant 3 cover
0.5 million ha; and 13 megasites in low irreplaceability–
low vulnerability quadrant 4 cover 0.8 million ha.
Our grouping of megasites into quadrants (Fig. 4) dif-
fers from that of Margules and Pressey (2000) in that we
give slightly higher priority to the upper left quadrant
(our quadrant 2, their quadrant 3) over the lower right
quadrant. We believe that sites of high and irreplaceable
biological value merit conservation action even if not
highly threatened today. Protecting these sites while
they are still reasonably intact is sensible. The private
lands in these areas are generally less expensive to pro-
tect than those in more threatened sites because they
are usually in areas with lower population and develop-
ment pressure.
Our proposed portfolio, if protected in its entirety,
would cover 43% more of the region than the current re-
serve network. For that 43% increment, there is a consider-
able “bang for the buck” for many elements—for example,
a 71% increase (to 100%) in coverage of G1/G2 species, a
62% increase for all special elements combined, and a 50%
increase for representation of ecological systems.
Progress toward conservation goals could be achieved
most effectively by protecting first the highest-priority
megasites (quadrant 1), then the medium-priority me-
gasites (quadrants 2 and 3), and finally the lower-priority
megasites (quadrant 4), (Table 2). The greatest gains
would be achieved with protection of the 15 highest-pri-
ority megasites (quadrant 1), resulting in an overall in-
crease in goal achievement of 30% for the three tracks
(from 47% of goals achieved currently to 77%), with an
increase in reserved area of 22%. Protection of known
locations of the most highly imperiled species and repre-
sentation of geoclimatic classes would increase by 46%
and 49%, respectively. The increment for focal species is
much less—only a 10% increase from adding quadrant 1
megasites to the reserve network—because most of the
best (e.g., roadless) habitat is already in parks and wil-
derness areas.
Figure 3. Distribution and demographic potential of grizzly bears in the expanded study region under (a) current
and (b) future landscape conditions, assuming road development on both private and public lands. Legend shows
population growth rate (lambda) values predicted by the PATCH model simulations. Hexagons represent individ-
ual territories.
904
Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
Protecting the 10 megasites from quadrant 2 would in-
crease overall achievement of goals from the three ap-
proaches another 11% to 88%. Protecting the 5 me-
gasites in quadrant 3 would increase achievement to
91%, and protecting the 13 megasites in quadrant 4
would increase achievement to 94%. Thus, our pro-
posed portfolio would encompass over 94% of special-
element occurrences, focal species habitat, and terres-
trial and aquatic ecological systems, at targeted levels,
within the GYE.
Conclusions
Our methodology is noteworthy in that it combines sev-
eral approaches and techniques that have previously
been applied separately in conservation planning. For
example, to our knowledge this is the first case where
spatially explicit modeling of population viability has been
combined with comprehensive representation analysis.
Hence, we were able to address questions concerning
the adequacy of alternative reserve networks that are ig-
nored in most reserve designs based on site-selection al-
gorithms (Possingham et al. 2000; Pressey et al. 2000).
Our results provide a reasonably complete assessment
of regional-scale conservation opportunities in the Greater
Yellowstone Ecosystem, given available information. Like
other researchers (e.g., Leslie et al. 2002), we were
pleased with the ability of the simulated annealing algo-
rithm to portray multiple conservation scenarios. These
scenarios, and the goals upon which they are based, can
be evaluated and refined by decision-makers through a
flexible, interactive process. Hence, the algorithm is a
decision-support tool, not a “black box” that yields a sin-
gle best solution. Nevertheless, our assessment is a snap-
shot of the existing conservation scene in the GYE, aug-
mented (for focal species) by a projection of current
trends into the future. To be most useful, new assess-
ments of the irreplaceability and vulnerability of sites
should be conducted periodically as particular threaten-
ing processes increase or diminish and as sites are added
to the reserve system or lost to development (Pressey &
Taffs 2001). In the real world, conservation priorities
change continuously.
We suspect that the three classes of ecological sys-
tems—vegetation, geoclimatic types, and aquatic habi-
tats—represented by our portfolio provide a functional
coarse filter (Noss 1987), although the coarse-filter hy-
pothesis cannot be tested rigorously without a complete
inventory of a region’s biota. More uncertainty is associ-
ated with special elements. Although the GYE, because
of its popularity with naturalists, has been better sur-
veyed than most regions of the American West, biases in
heritage-program databases are inevitable. For instance,
some portions of the GYE are poorly known biologi-
cally, such that the absence of element occurrences
Figure 4. Irreplaceability versus
vulnerability of megasites based on
nine conservation criteria. Mega-
sites are grouped by priority quad-
rants 1–4.
Conservation Biology
Volume 16, No. 4, August 2002
Noss et al. Irreplaceability and Vulnerability of Greater Yellowstone Sites
905
from these areas probably reflects an absence of surveys
more than an absence of imperiled species and commu-
nities. Heritage databases do not allow these two classes
of absence to be discriminated. With more complete
field surveys, some megasites with low irreplaceability
scores might move upward in priority.
Our treatment of focal species differs from conven-
tional approaches in that we use both static and dy-
namic models to evaluate the relative demographic
value of sites. The carnivore populations in the GYE
are on the periphery of their ranges due to climatic or
historical factors and cannot expect a large demo-
graphic rescue effect from surrounding regions. Cur-
rently, the core refugia of the national parks and adja-
cent wilderness areas are strong sources and could
support carnivore populations in the extremities of the
region. If current trends continue, development will in-
creasingly surround the core GYE with sinks, weakening
its ability to sustain populations in outlying areas. It is
noteworthy that areas of high value for multiple species
combine high biological productivity and security from
human impacts (Carroll et al. 2001). Such areas (e.g., un-
developed riparian areas) are scarce in the GYE and tend
to be highly threatened by development (Hansen et al.
1999).
Although we have used such terms as “protected ar-
eas” and “reserves” to describe the desired fate for mega-
sites, conservation objectives can be met through a
variety of means. Protection and management options
include direct fee acquisition, conservation easements,
management agreements and stewardship assistance to
landowners, agency designations of special areas (e.g.,
research natural areas), congressional wilderness desig-
nations, and administrative actions such as national mon-
Figure 5. Megasites ( by quadrant)
and protected areas of the Greater
Yellowstone Ecosystem.
906
Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
ument designations. Our simulations of the effects of al-
ternative future scenarios on the viability of focal
species showed that the straightforward action of avoid-
ing new road building in existing roadless areas on pub-
lic lands would have highly positive results (Noss et al.
2001). In keeping with the precautionary principle, the
highest-priority (e.g., quadrant 1) megasites generally
should receive the highest level of protection, whereas
lower-priority megasites could accommodate more hu-
man uses.
Protection opportunities in the GYE will not arise in
an orderly sequence that corresponds to science-based
priorities. For example, megasites in quadrant 3 may be-
come available for protection before megasites in quad-
rant 1. If not protected quickly, some of these sites may
be converted to subdivisions. Yet funds, or political cap-
ital, spent protecting these sites may preclude opportu-
nities for protecting biologically more significant sites in
the future. What is the optimal course of action under
such circumstances?
The irreplaceability-vulnerability approach to recog-
nizing conservation priorities is not perfectly suited to
real-world opportunities (Pressey & Taffs 2001). We sug-
gest that conservationists follow an informed opportun-
ism, taking advantage of conservation openings as they
arise, but with explicit recognition of the trade-offs in-
volved. Systematic conservation planning allows the ef-
fects of single and cumulative decisions to be quantified
and considered in a biologically meaningful way (Mar-
gules & Pressey 2000). With information made transpar-
ent and explicit, decision-makers will be better equipped
to take actions that are scientifically defensible and that
result in the most biodiversity conserved.
Acknowledgments
This work was funded by the Greater Yellowstone Coali-
tion with assistance from The Nature Conservancy, the
Doris Duke Foundation, and others. M. Lammert of The
Nature Conservancy constructed the aquatic-habitat clas-
sification system and provided helpful advice on inte-
grating it into our study. M. Scott and D. Glick of the
Greater Yellowstone Coalition made this study possible
and helped us see it through to completion. We thank R.
Pressey, H. Possingham, and an anonymous reviewer for
the most detailed and constructive comments we have
ever received on a manuscript.
Literature Cited
Andelman, S., I. Ball, F. Davis, and D. Stoms. 1999. SITES V 1.0: an ana-
lytical toolbox for designing ecoregional conservation portfolios.
The Nature Conservancy, Boise, Idaho.
Boyce, M. S., and L. L. McDonald. 1999. Relating populations to habi-
tats using resource selection functions. Trends in Ecology & Evolu-
tion
14:
268–272.
Briers, R. A. 2002. Incorporating connectivity into reserve selection
procedures. Biological Conservation
103:
77–83.
Carroll, C., R. F. Noss, and P. C. Paquet. 2001. Carnivores as focal spe-
cies for conservation planning in the Rocky Mountain region. Eco-
logical Applications
11:
961–980.
Clark, T. W., S. C. Minta, A. P. Curlee, and P. M. Kareiva. 1999. A model
Table 2. Achievement of conservation goals in the Greater Yellowstone Ecosystem, beginning with the current protected-areas network and
continuing with addition of megasites in priority quadrants 1–4.
Current
protected
areas (%)
Current protected areas plus (%) Total change
(%)quad 1 quad 2 quad 3 quad 4
Protected area 26.6 48.4 58.2 62.5 69.8
43.2
Special elements
all G1–G2* 28.9 74.9 89.1 93.3 100
71.1
class 1, local-scale species 40.7 69.2 86.3 89.0 93.2
52.5
class 2, birds 26.0 67.4 80.3 83.7 85.6
59.6
class 4, fish 26.7 55.6 82.2 84.4 87.8
61.1
class 5, plant communities 28.7 81.6 91.7 94.8 95.0
66.3
Special-elements average 30.2 69.7 85.9 89.0 92.3
62.1
Focal-species habitat
grizzly bear 94.4 96.5 98.0 98.3 98.9
4.5
wolf 77.8 86.0 92.7 94.1 96.3 18.5
wolverine 41.3 62.7 74.2 76.0 83.1 41.8
Focal-species average 71.2 81.7 88.3 89.5 92.8 21.6
Representation (ecological systems)
15% vegetation types 61.5 89.7 92.3 92.3 100 38.5
15% geoclimatic classes 41.0 89.7 92.3 94.9 100 59.0
20% aquatic types 44.3 74.5 90.5 95.9 98.0 53.7
Representation average 48.9 84.6 91.7 94.4 99.3 50.4
Total average 46.5 77.1 88.1 90.6 94.4 47.9
*G1, critically imperiled globally; G2, imperiled globally.
Conservation Biology
Volume 16, No. 4, August 2002
Noss et al. Irreplaceability and Vulnerability of Greater Yellowstone Sites 907
ecosystem for carnivores in Greater Yellowstone. Pages 1–9 in
T. W. Clark, A. P. Curlee, S. C. Minta, and P. M. Kareiva, editors.
Carnivores in ecosystems: the Yellowstone experience. Yale Uni-
versity Press, New Haven, Connecticut.
Craighead, F. C., Jr. 1979. Track of the grizzly. Sierra Club Books, San
Francisco, California.
Daly, C., R. P. Neilson, and D. L. Phillips. 1994. A statistical-topo-
graphic model for mapping climatological precipitation over
mountainous terrain. Journal of Applied Meteorology 33:140–
158.
Dinerstein, E., D. M. Olson, D. H. Graham, A. L. Webster, S. A. Pimm,
M. P. Bookbinder, and G. Ledec. 1995. A conservation assessment
of the terrestrial ecoregions of Latin America and the Caribbean.
World Wildlife Fund and the World Bank, Washington, D.C.
Ferrier, S., R. L. Pressey, and T. W. Barrett. 2000. A new predictor of
the irreplaceability of areas for achieving a conservation goal, its
application to real-world planning, and a research agenda for fur-
ther refinement. Biological Conservation 93:303–325.
Groves, C., L. Valutis, D. Vosick, B. Neely, K. Wheaton, J. Touval, and
B. Runnels. 2000. Designing a geography of hope: a practitioner’s
handbook for ecoregional conservation planning. The Nature Con-
servancy, Arlington, Virginia.
Hansen, A. J., J. J. Rotella, M. P. V. Kraska, and D. Brown. 1999. Dy-
namic habitat and population analysis: an approach to resolve the
biodiversity manager’s dilemma. Ecological Applications 9:1459–
1476.
Kirkpatrick, J. B., and M. J. Brown. 1994. A comparison of direct and
environmental domain approaches to planning reservation of for-
est higher plant communities and species in Tasmania. Conserva-
tion Biology 8:217–224.
Knight, D. H. 1994. Mountains and plains: the ecology of Wyoming
landscapes. Yale University Press, New Haven, Connecticut.
Lambeck, R. J. 1997. Focal species: a multi-species umbrella for nature
conservation. Conservation Biology 11:849–856.
Leslie, H., M. Ruckelshaus, I. R. Ball, S. Andelman, and H. P. Possing-
ham. 2002. Using siting algorithms in the design of marine reserve
networks. Ecological Applications: in press.
Manly, B. F. J., L. L. McDonald, and D. L. Thomas. 1993. Resource se-
lection by animals. Chapman and Hall, New York.
Margules, C. R., and R. L. Pressey. 2000. Systematic conservation plan-
ning. Nature 405:243–253.
McDonnell, M., H. P. Possingham, I. R. Ball, and E. Cousins. 2002.
Mathematical methods for spatially cohesive reserve design. Envi-
ronmental Modelling and Assessment: in press.
Merrill, T., D. J. Mattson, R. G. Wright, and H. B. Quigley. 1999. Defin-
ing landscapes suitable for restoration of grizzly bears Ursus arctos
in Idaho. Biological Conservation 87:231–248.
Miller, B., R. Reading, J. Strittholt, C. Carroll, R. Noss, M. Soulé, O.
Sanchez, J. Terborgh, D. Brightsmith, T. Cheeseman, and D. Fore-
man. 1998–1999. Using focal species in the design of nature re-
serve networks. Wild Earth 8(4):81–92.
Mladenoff, D. J., T. A. Sickley, R. G. Haight, and A. P. Wydeven. 1995.
A regional landscape analysis and prediction of favorable gray wolf
habitat in the northern Great Lakes region. Conservation Biology 9:
279–294.
Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. da Fonseca, and
J. Kent. 2000. Biodiversity hotspots for conservation priorities. Na-
ture 403:853–858.
Noss, R. F. 1983. A regional landscape approach to maintain diversity.
BioScience 33:700–706.
Noss, R. F. 1987. From plant communities to landscapes in conserva-
tion inventories: a look at The Nature Conservancy (USA). Biologi-
cal Conservation 41:11–37.
Noss, R. F. 2001. Beyond Kyoto: forest management in a time of rapid
climate change. Conservation Biology 15:578–590.
Noss, R. F., and A. Cooperrider. 1994. Saving nature’s legacy: protect-
ing and restoring biodiversity. Island Press, Washington, D.C.
Noss, R. F., H. B. Quigley, M. G. Hornocker, T. Merrill, and P. C.
Paquet. 1996. Conservation biology and carnivore conservation in
the Rocky Mountains. Conservation Biology 10:949–963.
Noss, R. F., M. A. O’Connell, and D. D. Murphy. 1997. The science of
conservation planning: habitat conservation under the Endangered
Species Act. Island Press, Washington, D.C.
Noss, R. F., J. R. Strittholt, K. Vance-Borland, C. Carroll, and P. Frost.
1999. A conservation plan for the Klamath-Siskiyou ecoregion. Nat-
ural Areas Journal 19:392–411.
Noss, R. F., G. Wuerthner, K. Vance-Borland, and C. Carroll. 2001. A
biological conservation assessment for the greater Yellowstone ec-
osystem. Report to the Greater Yellowstone Coalition, Bozeman,
Montana. Conservation Science, Corvallis, Oregon. Available from
www.conservationscience.com (accessed December 2001).
Patten, D. T. 1998. Riparian ecosystems of semi-arid North America: di-
versity and human impacts. Wetlands 18:498–512.
Poiani, K. A., B. D. Richter, M. G. Anderson, and H. E. Richter. 2000.
Biodiversity conservation at multiple scales: functional sites, land-
scapes, and networks. BioScience 50:133–146.
Possingham, H. P., I. R. Ball, and S. Andelman. 2000. Mathematical
methods for identifying representative reserve networks. Pages
291–306 in S. Ferson and M. Burgman, editors. Quantitative meth-
ods for conservation biology. Springer-Verlag, New York.
Pressey, R. L. 1994. Ad hoc reservations: forward or backward steps in
developing representative reserve systems. Conservation Biology
8:662–668.
Pressey, R. L., and R. M. Cowling. 2001. Reserve selection algorithms
and the real world. Conservation Biology 15:275–277.
Pressey, R. L., and K. H. Taffs. 2001. Scheduling conservation action in
production landscapes: priority areas in western New South Wales
defined by irreplaceability and vulnerability to vegetation loss. Bio-
logical Conservation 100:355–376.
Pressey, R. L., C. J. Humphries, C. R. Margules, R. I. Vane-Wright,
and P. H. Williams. 1993. Beyond opportunism: key principles
for systematic reserve selection. Trends in Ecology & Evolution
8:124–128.
Pressey, R. L., I. R. Johnson, and P. D. Wilson. 1994. Shades of irre-
placeability: towards a measure of the contribution of sites to a res-
ervation goal. Biodiversity and Conservation 3:242–262.
Pressey, R. L., T. C. Hager, K. M. Ryan, J. Schwarz, S. Wall, S. Ferrier,
and P. M. Creaser. 2000. Using abiotic data for conservation assess-
ments over extensive regions: quantitative methods applied across
New South Wales, Australia. Biological Conservation 96:55–82.
Ricketts, T. H., E. Dinerstein, D. M. Olson, C. J. Loucks, W. M. Eich-
baum, D. A. DellaSala, K. C. Kavanagh, P. Hedao, P. T. Hurley, K. M.
Carney, R. A. Abell, and S. Walters. 1999. A conservation assess-
ment of the terrestrial ecoregions of North America. I. The United
States and Canada. Island Press, Washington, D.C.
Schumaker, N. H. 1998. A user’s guide to the PATCH model. EPA/600/
R-98/135. U.S. Environmental Protection Agency, Environmental
Research Laboratory, Corvallis, Oregon.
Scott, J. M., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves, J.
Anderson, S. Caicco, F. D’Erchia, T. C. Edwards, J. Ulliman, and
R. G. Wright. 1993. Gap analysis: a geographical approach to pro-
tection of biological diversity. Wildlife Monographs 123.
Scott, J. M., T. H. Tear, and F. W. Davis. 1996. Gap analysis: a land-
scape approach to biodiversity planning. American Society for Pho-
togrammetry and Remote Sensing, Bethesda, Maryland.
Scott, J. M., F. W. Davis, G. McGhie, R. G. Wright, C. Groves, and J.
Estes. 2001. Nature reserves: do they capture the full range of
America’s biological diversity? Ecological Applications 11:999–
1007.
Soil Survey Staff. 1992. State soil geographic data base (STATSGO) data
user’s guide. Miscellaneous publication 1492. U.S. Department of
Agriculture Soil Conservation Service, U.S. Government Printing Of-
fice, Washington, D.C.
Stein, B. A., and F. W. Davis. 2000. Discovering life in America: tools
908 Irreplaceability and Vulnerability of Greater Yellowstone Sites Noss et al.
Conservation Biology
Volume 16, No. 4, August 2002
and techniques of biodiversity inventory. Pages 19–53 in B. A.
Stein, L. S. Kutner, and J. S. Adams, editors. Precious heritage: the
status of biodiversity in the United States. Oxford University Press,
Oxford, United Kingdom.
Theobald, D. M. 2001. Technical description of mapping historical,
current, and future housing densities in the US using census block
groups. Colorado State University, Fort Collins. Available from
http://ndis.nrel.colostate.edu/davet/dev patterns.htm (accessed De-
cember 2001).
United States Geological Survey (USGS). 2001. Delineation and devel-
opment of digital boundary data for fifth- and sixth-level hydrologic
units: pilot program for northeast area of Wyoming. USGS, Chey-
enne, Wyoming. Available from http://wy.water.usgs.gov/projects/
watershed/index.htm (accessed December 2001).
