Recovering Riparian Plant Communities with

Wolves in Northern Yellowstone, U.S.A.

Robert L. Beschta1,2 and William J. Ripple1

Abstract transects, heights of woody species have begun to exceed

Gray wolves (Canis lupus) were extirpated from Yellow- 200 cm (the approximate upper browse level of Elk). Re-
stone National Park in the 1920s. The ensuing seven dec- sults are consistent with the reestablishment of a tri-level
ades marked a period when wild ungulates, principally trophic cascade involving wolves, ungulates, and riparian
Elk (Cervus elaphus), extensively used woody browse vegetation. We additionally present conceptual models
species in the upper Gallatin and northern winter ranges, of vegetation recovery, illustrating differences in plant
thus limiting the capability of establishing plants to grow height responses to behaviorally and density-mediated
more than 100 cm in height. Following the reintroduction trophic cascades. Northern Yellowstone’s ‘‘experiment in
of wolves in the mid-1990s, we evaluated patterns of time,’’ whereby wolves were removed and then reintro-
woody browse species recovery within riparian areas of duced, provides new insights regarding how top preda-
these winter ranges. Measurements indicated that cotton- tors can influence the structure and biodiversity of
wood (Populus spp.) recruitment (growth of seedlings/ terrestrial ecosystems. Restoration ecologists and policy-
sprouts into tall saplings and trees) was occurring for the makers should consider the potential benefits of large
first time in several decades. A spatially patchy increase predators as an option for helping restore degraded eco-
in the heights of young willow (Salix sp.) and cotton- systems.
wood in the upper Gallatin and northern winter ranges,
respectively, was also found within riparian transects Key words: restoration, riparian ecosystems, trophic
comprising nearly 20 km in total length. Within some cascades, ungulates Yellowstone, wolves.

Introduction ces. In contrast, annual herbivory by ungulates creates

Ecological restoration can include both active and passive a ‘‘press’’ disturbance regime. Although the annual incre-
approaches for reestablishing historical plant and animal mental cropping of vegetation may appear to be less sig-
communities, as well as the renewal of ecosystem func- nificant than the effects of pulse disturbances, if such
tions necessary to sustain those communities (Kauffman herbivory comprises a sufficiently large percentage of cur-
et al. 1997). Active restoration, for example, might range rent annual growth and is persistent (occurs year after
from minimizing human interventions to the reintroduc- year), these disturbances can have major impacts on plant
tion of previously extirpated species, thereby allowing nat- communities and ecosystems (Ohmart 1996; Belsky et al.
ural disturbances (passive restoration) to reorganize and 1999; Barmore 2003). Wolf predation throughout the year
restore an ecosystem over time. Inherent in any restora- also represents a press disturbance and its effects on con-
tion effort is an adequate understanding of how ecosys- sumers (ungulates) may be transmitted to plant communi-
tems historically functioned (reference conditions) and ties as a trophic cascade (Estes et al. 2001). Over time,
how they have changed over time (Ripple & Beschta and in conjunction with other predators, this wide-ranging
2005). Such information is usually crucial for assessing the carnivore may influence not only the size of prey popula-
magnitude of historical and contemporary human effects, tions (density mediation) but also their patterns of herbiv-
as well as establishing recovery goals. ory (behavioral mediation).
Disturbances alter the function, species composition, During the 1800s and early 1900s, Euro-Americans erad-
and population structure of ecosystems. Wildfires and icated Gray wolves (Canis lupus) and Grizzly bears (Ursus
large floods represent ‘‘pulse’’ disturbances (Bender et al. arctos) from some 95 to 99% of their original range in the
1984) because individual events can reset plant communi- conterminous United States, thus allowing (1) the relaxa-
ties, often with extended time periods between occurren- tion of predation as a selective force and (2) the irruption
of herbivore populations (Berger 1999). Wolves, a large
predator capable of affecting a variety of animal species
1
College of Forestry, Oregon State University, Corvallis, OR 97331, U.S.A.
(Soulé et al. 2005), were extirpated from northern Yellow-
2
Address correspondence to R. L. Beschta, email robert.beschta@oregonstate. stone’s winter ranges in the early 1900s. During subsequent
edu decades, the effects of wild ungulate herbivory, principally
Ó 2008 Society for Ecological Restoration International
Elk (Cervus elaphus), upon riparian plant communities
doi: 10.1111/j.1526-100X.2008.00450.x remained a controversial management and scientific issue

380 Restoration Ecology Vol. 18, No. 3, pp. 380–389 MAY 2010
Riparian Recovery with Wolves

(e.g., Chadde and Kay 1991; NRC 2002; Wagner et al. Although YNP was established in 1872, market hunting
2006). However, most of this debate occurred without con- of ungulates and predators continued until 1886 when the
sidering the potential ecological role of wolves. U.S. Army was assigned responsibility for protecting the
The general thrust of this article embraces a major park’s wildlife (YNP 1997). However, persecution of
premise—following the extirpation of Gray wolves (an wolves continued inside and outside the park through the
apex predator), riparian plant communities within the early 1900s. With the extirpation of wolves in the mid-
winter ranges of northern Yellowstone were severely 1920s (Schullery & Whittlesey 1992), northern Yellow-
impacted by ungulate herbivory over a period of seven stone’s Elk had unimpeded access to winter range plant
decades (mid-1920s to mid-1990s). This premise is briefly communities, and browsing impacts to vegetation were
summarized in our description of study areas below. How- soon observed (Callahan 1923). In the upper Gallatin win-
ever, the primary goal of this study was to assess the ter range, Elk numbers generally decreased in subsequent
extent that trophic cascades following the reintroduction decades due to annual harvest (by hunting) of animals that
of wolves in the mid-1990s may have initiated recovery of migrated outside the park as well as periodic die-offs asso-
these plant communities. If removal, by humans, of ciated with a degraded winter range and the occurrence of
a wolf-dominated disturbance regime allowed ungulate severe winters (Peek et al. 1967). Annual hunting of Elk
herbivory to significantly impact winter range riparian outside the park’s portion of the northern range also
vegetation over multiple decades, then reinstatement of occurred. In addition, in the 1920s, the Park Service began
that disturbance regime may represent a prerequisite for capturing Elk within the northern range and shipping
initiating recovery of plant communities. them to other locations, both to assist in the reestablish-
ment of Elk herds in various parts of the western United
States and Canada and to reduce browsing impacts. Even-
tually, the agency resorted to killing Elk within the park
Study Areas in an attempt to control their impacts upon vegetation
Roughly 80 km apart along the northern portion of Yel- and soils.
lowstone National Park (YNP), the ‘‘upper Gallatin’’ and Even though reduced Elk numbers occurred during the
‘‘northern’’ winter ranges comprise approximately 200 and decades following wolf extirpation in the upper Gallatin
1,500 km2, respectively, of lower mountain slopes and val- winter range, Lovaas (1970) indicated that upland plant
ley bottom terrain in the northern Rocky Mountains communities were heavily grazed, Aspen ramets (root
(Barmore 2003; Ripple & Beschta 2004b). Most of the sprouts) were no longer able to grow above the browse
upper Gallatin winter range occurs outside the park, level of Elk, and hillslope erosion became a concern.
whereas approximately two-thirds of the northern winter Riparian willow communities were also heavily browsed
range occurs inside the park. Summers are short and cool, (Patten 1968; Ripple & Beschta 2004b), eventually leading
whereas winters are typically long and cold. Winter snow- to unstable channels and a hydrologically disconnected
packs are usually shallow along valley bottoms (elevation floodplain (Beschta & Ripple 2006).
approximately 2,000 m) and south-facing slopes, with Following the loss of wolves in the northern range, pal-
increasing snowpack depths at higher elevations. atable woody species were increasingly unable to establish
Low-elevation portions of these winter ranges encom- and grow above the browse level of Elk (NRC 2002;
pass shrub-steppe plant communities dominated by Big Barmore 2003). Similar to the upper Gallatin, once plant
sagebrush (Atremisia tridentata), with patches of inter- communities had been degraded, even reduced numbers
mixed Douglas fir (Pseudotsuga menziesii), Lodgepole of Elk were capable of continuing to suppress vegetative
pine (Pinus contorta), and Aspen (Populus tremuloides). growth. The Park Service’s Elk culling program in the
Riparian areas along the upper Gallatin River are domi- northern range was terminated in 1968, and within two
nated by willows (Salix spp.), whereas cottonwoods (Pop- decades, its Elk population of approximately 4,000 had
ulus spp.), willows, and Aspen variously occur within irrupted to nearly 19,000 animals. This large Elk popula-
riparian areas of Rose Creek, Soda Butte Creek, and the tion only served to increase the severity of impacts to
Lamar River in the northern range. plant communities and to curtail the recruitment (i.e.,
In addition to Gray wolves, Yellowstone’s large preda- growth of seedlings/sprouts into tall saplings or trees) of
tors include Grizzly bear, Black bear (Ursus americanus), woody browse species (Chadde & Kay 1991; Ripple &
and Cougar (Felis concolor). Elk frequent both winter Larsen 2000; Beschta 2005). Northern range vegetation
ranges along with smaller populations of Mule deer (Odo- studies that spanned over five decades (1935–1989) found
coileus hemionus), White-tailed deer (Od. virginianus), that average heights of young willow, Aspen, and other
Moose (Alces alces), Pronghorn (Antilocapra americana), woody browse species outside ungulate exclosures never
and Bighorn sheep (Ovis canadensis). Although Bison exceeded 83 cm (n ¼ 81 species-years), whereas the same
(Bison bison) have been historically absent from the species inside exclosures rapidly increased in height
upper Gallatin, a herd has been present in the northern (Chadde & Kay 1991; Singer 1996; Barmore 2003). Fruit
range throughout the 1900s. In recent years, the size of production of heavily browsed berry-producing shrubs
this herd has been increasing. outside exclosures was severely reduced or eliminated

MAY 2010 Restoration Ecology 381

 Riparian Recovery with Wolves

(Kay 1995). With the deterioration of woody plant com- heights of the three tallest Booth willows within each 100-
munities, Beaver (Castor canadensis) underwent decline, m segment of the same 3-km transect. We used linear
and by the early 1950s, only scattered colonies remained regression to illustrate trends in springtime willow heights,
(Jonas 1955). by year, along the 3-km transect. Using the linear relation-
During March 1995, 14 wolves were released into the ship of 2001 as a reference, we used multiple regression to
park with an additional 17 wolves the following winter. test for significant (p  0.05) slope and intercept differen-
They were soon breeding and establishing packs across ces in subsequent years.
northern Yellowstone. In the northern range study area, we measured late-
summer heights of young cottonwoods along riparian
transects paralleling major streams, including:
Methods
(1) ‘‘Lower Lamar’’—a 9.9-km transect (measured in
We summarized recent population estimates for wolves,
August 2002 and 2006) along the Lamar River down-
Elk, and Bison in the upper Gallatin and northern winter
stream from the Soda Butte Creek confluence.
ranges. In the northern range study area, we determined
(2) ‘‘Rose Creek’’—a 2.2-km transect (measured in
long-term patterns of cottonwood recruitment. A diame-
August 2002 and 2006) encompassing three distribu-
ter at breast height (dbh) of 5 cm was chosen as a threshold
taries of Rose Creek that flow across the alluvial fan
diameter for measurement as it was assumed that cotton-
at the Lamar Ranger Station (Buffalo Ranch).
woods of this diameter, and larger, would have attained
(3) ‘‘Soda Butte Creek’’—a 2.2-km transect (measured in
a height well above the browse level of Elk. We thus
August 2006) along Soda Butte Creek starting approx-
searched for and measured the diameters of all cotton-
imately 4-km upstream of its confluence with the
wood trees more than or equal to 5 cm in dbh during late
Lamar River.
summer of 2001 with reinventories in 2002, 2003, 2004,
(4) ‘‘Upper Lamar’’—a 1.8-km transect (measured in
and 2006. The establishment date of each inventoried cot-
August 2003 and 2006) along the Lamar River
tonwood was estimated from tree age versus dbh relation-
upstream of the Soda Butte Creek confluence.
ships to establish the age structure of cottonwoods in the
northern range study area (Beschta 2005). From those Within each 50-m segment of a transect, we searched
results, we regressed tree frequency versus establishment for the three tallest cottonwood seedlings or root sprouts
date for the decades when wolves were present (pre- more than or equal to 20 cm tall and averaged their
1920s). This exponential relationship provided a basis of heights. For evaluating height changes over time, we ex-
comparison regarding cottonwood tree recruitment during pressed segment heights of a transect as a percentile and
recent decades. compared any change in median height between sampling
We further characterized the general recovery status of periods. Height measurements of woody browse species in
riparian plant communities by measuring heights of young both the upper Gallatin and the northern winter ranges, in
Booth willow (Salix boothii) in the upper Gallatin and conjunction with other Yellowstone studies and the
young cottonwoods (Populus angustifolia and Po. tricho- broader literature, allowed us to develop conceptual mod-
carpa) in northern range study areas. We selected these els that demonstrate differences in plant community
woody species because they commonly occurred in their responses from behaviorally mediated (risk based) and
respective riparian areas, are highly palatable, and density-mediated (mortality based) components of a tro-
appeared to reflect the general pattern of height growth phic cascade.
for other woody browse species. Our sampling of plant
heights in riparian areas was undertaken to provide
a ‘‘leading edge’’ indication of ecosystem recovery in the Results
decade following wolf reintroduction.
In the upper Gallatin study area, we initially measured Wolves and Ungulates
Booth willow heights within a 3-km riparian transect along Following the 1995–1996 reintroduction of wolves, their
the Gallatin River in August 2003. This transect occurred numbers began to increase. Over the past 5 years, north-
along a predation risk gradient (low predation risk at the ern YNP populations have ranged between 54 and 106
upstream end of the transect, with increasing risk in wolves (Smith et al. 2006). Only two Elk population esti-
a downstream direction; Ripple & Beschta 2004b). Within mates, ranging from 1,048 to 3,028 animals, were available
each 100-m segment of the transect (30 segments in total), in the upper Gallatin study area for the 10 years (1987–
the three tallest Booth willow plants were selected. We 1996) leading up to wolf reintroduction. After wolf rein-
undertook plant architecture measurements (Keigley troduction, census estimates indicate that approximately
1998; Beschta & Ripple 2007) on the tallest stem (leader) 1,000 Elk have frequented this winter range (Fig. 1a). For
of each selected plant to evaluate its browsing history and the northern range, Elk populations peaked at approxi-
springtime height (after winter browsing) for the previous mately 19,000 animals in the decade prior to wolf reintro-
2 years. In August 2006, we measured springtime and late- duction. Elk densities after wolf reintroduction have
summer (after current annual growth had occurred) leader trended downward, whereas Bison densities have recently

382 Restoration Ecology MAY 2010

Riparian Recovery with Wolves

begun to increase (Fig. 1b). Although Bison numbers recent decades. During the 2002 and 2003 reinventories,
remain lower than those of Elk, it should be noted that we were unable to find any cottonwoods that had recently
Bison are considerably larger than Elk and often forage attained or exceeded the threshold dbh of 5 cm. However,
throughout the year within this winter range, whereas Elk in 2004, we found 19 cottonwoods more than or equal to 5
use is more seasonal (i.e., late fall to early spring). Thus, cm in dbh, and by 2006 (cottonwood stands were not
the foraging needs of the northern range Bison population inventoried in 2005), the count had increased to 76 cotton-
are considerably greater than a simple comparison of pop- woods (ranging between 5 and 11 cm in dbh). Cottonwood
ulation size between Bison and Elk would indicate. age structure over the past 200 years is presented in
Figure 2 for the northern range study area, exclusive of
data from the Rose Creek site where cottonwood recruit-
Riparian Vegetation
ment was historically influenced by Park Service culling
In 2001, a total of 954 cottonwoods more than or equal to operations (Beschta 2005). The fitted exponential rela-
5 cm in dbh were inventoried along the Lower Lamar, tionship, based on pre-1920 data when wolves were pres-
Upper Lamar, and Soda Butte Creek sites, of which 836 ent, provides a general basis for identifying expected
remained in 2006. This represented 12% mortality over levels of recruitment after 1920.
the 5-year monitoring period (equivalent to an annual For the 76 cottonwoods that had recently attained a dbh
compound mortality rate of 2.6% per year). If this rate of of more than or equal to 5 cm (with establishment dates in
overstory tree loss were to continue, the number of stand- the mid- to late-1990s; Fig. 2), 29 occurred along the Lower
ing trees across these sites will be halved within approxi- Lamar site, of which 23 were found on a mid-channel
mately 26 years. island (Fig. 3a), 5 were associated with a gully, and 1 was
The 2001 cottonwood inventory also found that trees physically protected from browsing by an accumulation of
between 5 and 15 cm in dbh were entirely absent, thus woody debris. However, large areas of bare alluvium
indicating that cottonwood recruitment had ceased during (potential germination sites for cottonwood seedlings)
along the Lower Lamar remain devoid of young cotton-
wood (Fig. 3b). Even where establishing cottonwoods (and
willows) were present, they were typically heavily browsed
and of short stature (<50 cm in height). Forty-seven cot-
tonwoods more than or equal to 5 cm in dbh were found
within the Upper Lamar site, typically near individual logs
and accumulations of logs, at the base of an eroding tall
terrace, or in association with the increasing heights of
other woody plants (e.g., willows or alder [Alnus sp.]).
Based on the exponential relationship of cottonwood fre-
quency versus establishment date for when wolves were
historically present (Fig. 2), the 76 cottonwoods more than
or equal to 5 cm in dbh that established in the 1990s repre-
sented only one-fourth of the number needed for sustain-
ing cottonwood populations at pre-1920 levels.
For the upper Gallatin study area, trends in willow
heights in recent years along the 3-km riparian transect
are illustrated in Figure 4. Linear regression of springtime
plant height versus transect distance, by year, indicated
that (1) willows were mostly less than 100 cm in height in
2001 and (2) height increases between 2001 and 2006
occurred mostly at the down-valley end of the transect. By
late summer of 2006, 50% of the transect segments had
average willow heights more than or equal to 200 cm
(Fig. 5a). A willow height of approximately 200 cm
appears to represent the normal upper browse level for
Figure 1. (a) Elk populations for the upper Gallatin winter range and Elk (Beschta & Ripple 2007).
(b) Elk and Bison populations for the northern winter range during For the northern range study area, riparian transects
the decade before and after wolf reintroduction, in YNP. Data sour- indicated major differences in temporal trends between
ces: (a) 1995–2003 Elk counts from Ripple and Beschta (2004b); sites based on the median (50th percentile) heights of
unpublished 2004–2006 Elk counts from Montana Department of
Fish, Wildlife, and Parks and (b) 1987–2004 Elk and Bison counts
young cottonwood. Here, the median cottonwood height
from White and Garrott (2005b); unpublished 2005–2006 Elk and for transect segments along the Lower Lamar remained
Bison counts from YNP. Elk counts for 1989, 1991, and 2006 likely 30 cm between 2002 and 2006, decreased from 70 to 52 cm
underestimate populations due to poor survey conditions. along Rose Creek between 2002 and 2006, and increased

MAY 2010 Restoration Ecology 383

 Riparian Recovery with Wolves

Figure 2. Frequency of cottonwood trees (Y) versus establishment date (X) within the northern range study area. The dashed line represents an
exponential relationship (Y ¼ [1.22 3 10211]e0.0154x, r2 ¼ 0.92, p < 0.01) based on decadal tree frequencies prior to 1920 (when wolves were
present); the symbol ‘‘*’’ denotes that the number of trees for a given decade is less than the lower 95% confidence limit of the exponential
regression line.

from 100 to 293 cm along the Upper Lamar between 2003 (3) The former extirpation of an apex predator has al-
and 2006 (Fig. 5b). No change or a decrease in median lowed large herbivores to maintain woody browse spe-
height over time indicates that herbivory is continuing to cies in a suppressed condition (low stature) across all
suppress the height growth of these young cottonwoods. sites (e.g., woody browse species in the upper Gallatin
In 2002, inspection of a 5-km reach of Soda Butte Creek and northern winter ranges after wolves had been
(Ripple and Beschta 2003) above its confluence with the eliminated).
Lamar River found only one location with young cotton- (4) The previously extirpated apex predator is reintro-
woods more than or equal to 200 cm in height. Cotton- duced at t ¼ 0, again completing the predator guild.
wood measurements along the 2.2 km Soda Butte Creek
Our first conceptual model of plant height responses
transect in 2006 indicated a median height of 170 cm
following reintroduction of an apex predator assumes
(Fig. 5b) and that 25% of the 50-m segments had heights
behaviorally mediated foraging patterns. In this model
(average of three tallest plants per 50-m segment) more
(Fig. 6a), herbivores begin to avoid sites of high predation
than or equal to 200 cm, thus indicating increased height
risk because of the presence of the reintroduced predator.
growth of young cottonwoods along this study reach in
Because predator avoidance represents learned behavior
recent years.
(based on predator–herbivore encounters), some time
may be required before browsing levels decrease on high-
Conceptual Models risk sites and plants to increase in height. Height increases
are initially greatest at high-risk sites, with moderate-risk
Based on our field results and observations, as well as
sites demonstrating a slower rate of increase. Low-risk
other published studies, we formulated conceptual models
sites may actually experience a general loss of plant height
of vegetation response following the reintroduction of
due to increased herbivory (e.g., Rose Creek site in the
a top predator. These models illustrate how heights of
northern range) as herbivores weigh foraging needs
establishing woody browse species may change over time
against predation concerns within a landscape of fear
(t) when behavioral or density mediation dominates a tro-
(Brown et al. 1999; Laundré et al. 2001; Ripple & Beschta
phic cascade. To assess the relative importance of these
2004a; Hernández & Laundré 2005; Berger 2007). This
two mechanisms, we considered three conceptual models
model indicates that the occurrence of plants increasing in
using plant height as an index of recovering conditions. In
height (high-risk sites) and those which are unable to
each model, we assumed the following initial conditions:
increase in height or are further suppressed by increased
(1) Site productivity is constant across all foraging sites. browsing (low-risk sites) can be attributed to behaviorally
(2) Foraging sites can be arrayed along a predation risk mediated foraging (Schmitz et al. 2000). Plant heights at
continuum (low to high). t ¼ n reflect a balance between current annual growth and

384 Restoration Ecology MAY 2010

Riparian Recovery with Wolves

Figure 4. Linear regressions of springtime Booth willow heights ver-

sus transect distance, by year, for a 3-km transect along the Gallatin
River in the upper Gallatin winter range (‘‘0 distance’’ is the
upstream end of the transect). Reintroduced wolves colonized the
upper Gallatin in the mid-1990s. There was strong evidence (F[3,94] ¼
4.57, p < 0.01) that the relationship between willow height (cm) and
distance (km) differed among years. Although there was no signifi-
cant difference in intercepts between 2001 and subsequent years
(p  0.54) or for slope between 2001 and 2002 (p ¼ 0.20), slopes for
the 2003 and 2006 relationships relative to 2001 were significantly
different (p < 0.01).

Figure 3. (a) A high predation risk mid-channel island surrounded by would be inversely related to herbivore density. This con-
significant escape impediments such as steep stream banks, woody ceptual model indicates that lower ungulate densities may
debris, coarse gravel surfaces, and variable river depths where woody be necessary before a more expansive recovery of riparian
browse species have been increasing in height following wolf reintro-
duction and (b) a low predation risk riparian area with good visibility
plant communities can occur in Yellowstone’s northern
and few escape impediments where establishing woody browse spe- ranges.
cies (e.g., cottonwood, willows) remain sparse, heavily browsed, and Our third conceptual model (Fig. 6c) combines the
of low stature (mostly <50 cm in height) after wolf reintroduction; behavioral and density mediation effects of a trophic
trees in center of the photo are mature cottonwoods and the ungu- cascade by simply averaging, over time, plant height
lates are Bison. Photos (a) and (b) are for late summer of 2003 and responses illustrated in the first two models. In this com-
2006, respectively; both photos show portions of the Lower Lamar
bined model, plants at high-risk sites experience the great-
study site.
est increase in height for t  1, although incremental
height increases also occur at sites of moderate and low
predation risk.
herbivory along a predation risk gradient. In this model
formulation, only plants on high-risk sites can eventually
grow above the upper foraging level of the prevalent
herbivores. Discussion
Our second conceptual model of recovering vegetation Other than humans, who occupy an ecological niche as
involves only density-mediated foraging. In this model top predator in most ecosystems, Peterson et al. (2003)
(Fig. 6b), we additionally assume that (1) herbivore densi- indicated that Gray wolves, by virtue of their widespread
ties have become sufficiently low at t ¼ 0 that plant height geographic distribution, group hunting, and year-round
increases begin to occur and (2) even though an apex activity, represent the most significant ungulate predator
predator was introduced at t ¼ 0, herbivore use of foraging in the Northern Hemisphere. If so, the loss or reintroduc-
sites occurs independently of predation risk. Thus, palat- tion of wolves may sufficiently alter trophic cascades
able plants begin to increase in height on all sites (t ¼ 1, 2) within a predator–consumer–producer system such that
and continue to grow above the upper foraging level of major adjustments ultimately occur at the lower most tro-
the prevalent herbivores (t  3). Eventually (t ¼ n), full phic level.
complements of height (and age) classes of riparian plants The winter ranges of northern Yellowstone represent
occur across all sites. Because site productivity was an experiment in time, whereby wolves were first extir-
assumed constant, the time required for plant heights to pated and then reintroduced. Although these treatments
exceed the normal upper foraging level of herbivores were unplanned (i.e., the Park Service did not

MAY 2010 Restoration Ecology 385

 Riparian Recovery with Wolves

Figure 5. Box and whisker plots of late-summer plant heights (in per-
centiles) by year and site for (a) willows in the upper Gallatin winter
range and (b) cottonwoods in the northern winter range; wolves were
reintroduced in northern Yellowstone during the mid-1990s. ‘‘Percen-
tiles’’ indicate the percentage of transect segments having a height
(average height of the three tallest plants per 50-m segment) greater
than the indicated height.

intentionally set out to manipulate wolf populations as an

experiment), they nevertheless provide a unique opportu-
nity to assess ungulate and plant community effects associ-
ated with the removal and return of a top predator. In
terrestrial systems, examples of large carnivore trophic
cascades have been uncommon (Borer et al. 2005).
After the extirpation of wolves in northern Yellow-
stone, Elk have generally been considered the principal
herbivore of concern in both winter ranges. However, in
the northern winter range, a decreasing Elk population
and an increasing Bison population during recent years
indicate that Bison may be having a more important role
affecting plant communities. We observed noticeable lev-
els of summertime foraging by Bison on young cotton-
woods and willows (Ripple & Beschta 2006) along the
Lower Lamar and Rose Creek sites. These observations
of summertime herbivory, suppressed cottonwood heights,
and the results of a radio-collared Elk study (Boyce et al. Figure 6. Conceptual models of woody plant height increases follow-
2003), indicating a relatively low probability of Elk occur- ing the reintroduction of an apex predator (see text for details). Mod-
rence along the Lower Lamar and Rose Creek sites in els represent (a) behaviorally mediated, (b) density-mediated, and
recent years, suggest that Bison herbivory may be sup- (c) combined behaviorally and density-mediated trophic cascades
over time (0  t  n). For simplicity, we assume linear relationships.
pressing woody browse species at these sites. Although
Bison represented less than 1% of wolf kills during 1995–
2000 (Smith et al. 2004), this had increased to 9% by 2005
(Smith et al. 2006), indicating that prey switching from the downstream end of the transect relative to the
Elk to Bison may be underway. upstream end. At the downstream end of this transect, the
Our results for the 3-km transect along the upper width of the valley narrows, multiple channels occur in the
Gallatin River indicate that willow heights were increas- floodplain, and a major highway comes relatively close to
ing along a spatial gradient several years following wolf the river. Thus, increased willow heights appear to be
reintroduction. In an analysis of 1998–2002 willow heights associated with terrain and cultural features that can affect
along this transect, Ripple and Beschta (2004b) found (1) the capability of Elk to detect and escape from wolves,
decreased browsing over time and (2) less browsing along that is, predation risk (Ripple & Beschta 2004a).

386 Restoration Ecology MAY 2010

Riparian Recovery with Wolves

For the northern range study area, increased median over time, particularly where multiple large predator and
heights of young cottonwoods along the Upper Lamar site herbivore species are present, lower trophic-level res-
were often associated with scattered cottonwood and coni- ponses of increased plant heights would likely assume a
fer logs that were relatively common along this reach in more complex pattern (e.g., the combined model, curvilin-
comparison to other riparian areas. In a few instances, ear relationships). Our results in the upper Gallatin indi-
these logs created local refugia by physically protecting cate height increases along a spatial gradient until 2003
woody browse species from ungulate herbivory. A tall ter- and more uniform height increases after 2003, perhaps
race (10–15 m in height) occurred along the eastern edge illustrating a shift from primarily a behaviorally mediated
of the floodplain, additionally limiting ungulate viewsheds trophic cascade to one where density mediation has
and escape routes, thus increasing predation risk. Simi- become more important.
larly, increases in the heights of young Aspen in recent In the northern range, a downward trend in Elk popula-
years have been found to be greatest on high-risk sites tions and an upward trend in Bison populations during the
(Ripple & Beschta 2007). In contrast, improvement in most recent decade complicate attempts to decipher the
median heights was not found for the Lower Lamar site relative importance of ungulate behavior versus ungulate
and an actual decrease in median heights occurred at the density upon recovering woody browse species. Neverthe-
Rose Creek site. Both sites occur in the widest part of the less, the occurrence of height increases (Soda Butte Creek
Lamar Valley where there is good visibility over long dis- and Upper Lamar sites) as well as no change or even
tances, woody debris accumulations are generally absent, a decrease in height (Lower Lamar and Rose Creek sites)
and other impediments to escape are relatively few and for young cottonwoods during the first decade of wolf
easily visible (low predation risk sites). recovery indicates that behavior mediation has likely been
Winter climates in northern YNP can influence large a major mechanism influencing plant communities in this
herbivore populations and seasonal patterns of herbivory winter range.
(e.g., deeper snowpacks tend to drive Elk to lower eleva- We focused on palatable deciduous woody species in
tions) (Boyce et al. 2003; White & Garrott 2005a). How- this study because of their fundamental importance to
ever, during the seven decades when wolves were absent riparian plant communities as well as their capacity to
from northern Yellowstone’s winter ranges and regardless index herbivore effects upon a broader range of plants
of climate fluctuations, Elk had relatively unconstrained and ecosystem processes over time. Prior to the extirpa-
access to riparian plant communities and woody browse tion of wolves, much evidence points to biologically di-
species were generally unable to grow above the browse verse riparian plant communities in Yellowstone’s northern
level of these ungulates. Now that wolves are back, how winter ranges. Yet, during the seven decades of wolf
variations in winter conditions (Vucetich et al. 2005) or absence, and regardless of ungulate densities or prevailing
long-term trends in climate might influence ungulate pop- climatic conditions, various studies found widespread deg-
ulations or patterns of herbivory in riparian areas remains radation of these plant communities. Following the rein-
to be seen. troduction of wolves in the mid-1990s, increased heights
Overall, our results, in concert with other recent studies of at least some willows, cottonwoods, and Aspen in Yel-
in the upper Gallatin and northern winter ranges, indicate lowstone’s northern winter ranges suggest that linkages
that willow, cottonwood, and Aspen recruitment is begin- between large predators, herbivores, and plants are again
ning to occur primarily within riparian areas where view- being reestablished. The recent return of beaver colonies
sheds, terrain, vegetation, or large wood features suggest to the northern range (Smith et al. 2003) serves to rein-
locally high predation risk, even though streamside areas in force this hypothesis. Nevertheless, numerous questions
general represent landscape conditions where Elk may be remain. Will increased heights of woody browse species
particularly vulnerable to wolf predation (Kunkel & continue to be associated primarily with high predation
Pletscher 2001; Gula 2004; Beyer 2006). The fact that cot- risk sites or will increases in heights become more wide-
tonwood recruitment is again underway in some northern spread in the coming years? How will changing Elk and
range riparian areas represents a fundamental departure Bison densities in the northern range study area, in combi-
from the downward trend in recruitment that occurred dur- nation with shifting environmental variables (e.g., climate
ing the 1920–1960s and the total absence of recruitment in change), influence future outcomes for plants? Large
the 1970s and 1980s, when wolves were absent. Neverthe- predator reintroduction is a relatively new occurrence in
less, contemporary levels of ungulate herbivory are continu- the United States, and the scientific basis for predicting
ing to suppress the growth of young cottonwoods, willows, expected ecological outcomes is limited. The extent to
and Aspen within many areas of the northern range. which plant height increases during the first decade of wolf
Our conceptual models indicate that spatially variable reintroduction will continue, or foreshadow a broader
increases in plant heights would be an expected outcome recovery of other native plant and animal species normally
of trophic cascades having a strong behavioral mediation associated with biologically diverse riparian ecosystems, is
component, whereas density mediation provides for yet to be determined.
recovery of plant communities across landscapes. How- Although these studies were conducted in the winter
ever, where large predator and herbivore densities change ranges of a national park, the occurrence of a trophic

MAY 2010 Restoration Ecology 387

 Riparian Recovery with Wolves

cascade involving Gray wolves may have implications to LITERATURE CITED

other areas of the northern Rocky Mountains (e.g., parts Barmore, W. J. 2003. Ecology of ungulates and their winter range in
of Idaho, Montana, and Wyoming) where reintroductions northern Yellowstone National Park: research and synthesis 1962-
have occurred. For other parts of the American West, 1970. Yellowstone Center for Resources, Yellowstone National
such as Oregon and Utah, dispersing wolves may eventu- Park, Wyoming, Montana.
Belsky, A. J., A. Matzke, and S. Uselman. 1999. Survey of livestock influ-
ally recolonize some of their former ranges (ODFW 2005;
ences on stream and riparian ecosystems in the western United
UDWR 2005). In the central Rocky Mountains, wolf rein- States. Journal of Soil and Waters Conservation 54:419–431.
troduction has been identified as a possible management Bender, E. A., T. J. Case, and M. E. Gilpin. 1984. Perturbation experi-
alternative to help reduce the adverse ecological effects of ments in community ecology: theory and practice. Ecology 65:1–13.
a burgeoning Elk population (RMNP 2006). Previously, Berger, J. 1999. Anthropogenic extinction of top carnivores and interspe-
Ratti et al. (2004) evaluated factors affecting the feasibil- cific animal behavior: implications of the rapid decoupling of a web
ity of reintroducing wolves to the Olympic Peninsula involving wolves, bears, moose, and ravens. Proceedings Royal Soci-
ety of London 266:2261–2267.
of Washington State but did not consider potential ecolog-
Berger, J. 2007. Carnivore repatriation and holarctic prey: narrowing the
ical effects wolves might have on lower trophic levels deficit in ecological effectiveness. Conservation Biology 21:
(i.e., plant communities). Although long-term studies by 1105–1116.
McLaren and Peterson (1994) at Isle Royale (a 544-km2 Beschta, R. L. 2005. Reduced cottonwood recruitment following extirpa-
island in Lake Superior) indicate that trophic interactions tion of wolves in Yellowstone’s northern range. Ecology 86:391–403.
have occurred between wolves, Moose (Alces alces), and Beschta, R. L., and W. J. Ripple. 2006. River channel dynamics following
extirpation of wolves in northeastern Yellowstone National Park,
Balsam fir (Abies balsamea), how these or other trophic
USA. Earth Surface Processes and Landforms 31:1525–1539.
cascades might relate to adjacent areas of Minnesota, Wis- Beschta, R. L., and W. J. Ripple. 2007. Increased willow heights along
consin, and Michigan where Gray wolves have recolonized northern Yellowstone’s Blacktail Deer Creek following wolf rein-
in recent years has been little explored by the scientific troduction. Western North American Naturalist 67:613–617.
community. In the Highlands of Scotland, where wolves Beyer, H. L. 2006. Wolves, elk and willow on Yellowstone National
have been absent since the mid-1700s and intensive forag- Park’s northern range. M.S. thesis. University of Alberta, Edmon-
ing by Red deer (Cervus elaphus) is a serious environmen- ton, Canada.
Borer, E. T., J. B. Seabloom, K. E. Shurnin, C. A. Anderson, B. Blanchetter,
tal problem, a recent assessment indicated that ungulate
and B. S. Halpern. 2005. What determines the strength of a trophic
densities could be significantly reduced through wolf rein- cascades? Ecology 86:528–537.
troduction (Nilsen et al. 2007). If some of the trophic-level Boyce, M. S., J. S. Mao, E. H. Merrill, D. Fortin, M. G. Turner, J. Fryxell,
effects measured in this study occur where wolves have and P. Turchin. 2003. Scale and heterogeneity in habitat selection by
been reintroduced or expanded their ranges, improved sus- elk in Yellowstone National Park. Ecoscience 10:421–431.
tainability and diversity of native flora and fauna may be Brown, J. S., J. W. Laundré, and M Gurung. 1999. The ecology of fear:
one of the important outcomes of such endeavors. optimal foraging, game theory, and trophic interactions. Journal of
Mammalogy 80:385–399.
Callahan, J. E. 1923. Annual elk and game report, spring 1923. USFS
Report. Gallatin National Forest, Bozeman, Montana.
Chadde, S. W., and C. E. Kay. 1991. Chapter 16. Tall-willow communities
Implications for Practice on Yellowstone’s winter range: a test of the ‘‘natural-regulation’’
d Annual herbivory by wild ungulates, in the absence paradigm. Pages 231–262 in R. B. Keiter and M. S. Boyce, editors.
The Greater Yellowstone ecosystem: redefining America’s wilder-
of an apex predator, can have profound long-term ef-
ness heritage. Yale University Press, New Haven, Connecticut.
fects on the composition, structure, and functioning Estes, J. A., K. Crooks, and R. Holt. 2001. Predators, ecological role of.
of riparian plant communities. Reducing these effects Pages 280–1 to 280–22 in S. Levin, editor. Encyclopedia of biodiver-
represents a high priority for restoring altered eco- sity, Vol. 4. Academic Press, San Diego, California.
systems. Gula, R. 2004. Influence of snow cover on wolf Canis lupus predation pat-
d The return of previously extirpated Gray wolves is terns in Bieszczady Mountains, Poland. Wildlife Biology 10:17–23.
allowing for passive restoration, via trophic cascades, Hernández, L., and J. W. Laundré. 2005. Foraging in the landscape of fear
and its implications for habitat use and diet quality of elk (Cervus
of some riparian plant communities in Yellowstone’s
elaphus) and bison (Bison bison). Wildlife Ecology 11:215–220.
northern winter ranges. Jonas, R. J. 1955. A population and ecological study of the beaver (Castor
d Managers and policymakers should consider the
canadensis) in Yellowstone National Park. Master’s thesis. Univer-
potential ecological benefits of large predators in sity of Idaho, Moscow, Canada.
efforts to recover/restore degraded ecosystems. Kauffman, J. B., R. L. Beschta, N. Otting, and D. Lytjen. 1997. An eco-
logical perspective of riparian and stream restoration in the Western
United States. Fisheries 22:12–24.
Kay, C. E. 1995. Browsing by native ungulates: effects on shrub and seed
Acknowledgment production in the greater Yellowstone ecosystem. Pages 310–320
in B. A. Roundy, E. D. McArthur, J. S. Haley, and D. K. Mann,
We appreciate the comments, questions, and suggestions compilers. Proceedings: wildland shrub and arid land restoration
of an earlier draft from C. Eisenberg, S. Lima, C. White, symposium. USDA Forest Service, General Technical Report INT-
and anonymous reviewers. GTR-315. Ogden, Utah.

388 Restoration Ecology MAY 2010

Riparian Recovery with Wolves

Keigley, R. B. 1998. Architecture of cottonwood as an index of browsing Ripple, W. J., and E. J. Larsen. 2000. Historic aspen recruitment, elk, and
history in Yellowstone. Intermountain Journal of Science 4:57–67. wolves in northern Yellowstone National Park, USA. Biological
Kunkel, K. E., and D. H. Pletscher. 2001. Winter hunting patterns of Conservation 95:361–370.
wolves in and near Glacier National Park, Montana. Journal of RMNP (Rocky Mountain National Park). 2006. Draft environmental
Wildlife Management 65:520–530. impact statement: elk and vegetation management plan. US Park
Laundré, J. W., L. Hernández, and K. B. Altendorf. 2001. Wolves, elk, Service, Rocky Mountain National Park, Estes Park, Colorado.
and bison: reestablishing the ‘‘landscape of fear’’ in Yellowstone Schmitz, O. J., P. A. Hambäck, and A. P. Beckerman. 2000. Trophic cas-
National Park, U.S.A. Canadian Journal of Zoology 792:1401–1409. cades in terrestrial systems: a review of the effects of carnivore re-
Lovaas, A. L. 1970. People and the Gallatin elk herd. Montana Fish and movals on plants. The American Naturalist 155:141–153.
Game Department, Helena. Schullery, P., and L. Whittlesey. 1992. The documentary record of wolves
McLaren, B. E., R. O. Peterson. 1994. Wolves, moose, and tree rings on and related wildlife species in Yellowstone National Park area prior
Isle Royale. Science 266:1555–1558. to 1882. Pages 1–174 in J. D. Varley and W. G. Brewster, editors.
Nilsen, E. G., E. J. Milner-Gulland, L. Schofield, A. Mysterud, N. C. Wolves for Yellowstone? A Report to the United States Congress,
Senseth, and T. Coulson. 2007. Wolf reintroduction to Scotland: volume IV: Research and analysis. National Park Service, Yellow-
public attitudes and consequences for red deer management. Pro- stone National Park, Mammoth Hot Springs, Wyoming, Montana.
ceedings of the Royal Society 274:995–1002. Singer, F. J. 1996. Differences between willow communities browsed by
NRC (National Research Council). 2002. Ecological dynamics on Yellow- elk and communities protected for 32 years in Yellowstone National
stone’s northern range. National Academy Press, Washington, D.C. Park. Pages 279–290 in F. J. Singer, editor. Effects of grazing by wild
ODFW (Oregon Department of Fisheries and Wildlife). 2005. Oregon ungulates in Yellowstone National Park. USDI, National Park Ser-
wolf conservation and management plan. ODFW, Salem, Oregon. vice, Natural Resource Information Division, Technical Report
Ohmart, R. D. 1996. Historical and present impacts of livestock grazing NPS/NRYELL/NRTR/96-01. Yellowstone National Park, Mam-
on fish and wildlife resources in western riparian habitats. Pages moth Hot Springs, Wyoming, Montana.
245–279 in P. R. Krausman, editor. Rangeland wildlife. Society for Smith, D. W., T. D. Drummer, K. M. Murphy, D. S. Guernsey, and S. B.
Range Management, Denver, Colorado. Evans. 2004. Winter prey selection and estimation of wolf kill rates
Patten, D. T. 1968. Dynamics of the shrub continuum along the Gallatin in Yellowstone National Park, 1995-2000. Journal of Wildlife Man-
River in Yellowstone National Park. Ecology 49:1107–1112. agement 68:153–166.
Peek, J. M., A. L. Lovaas, and R. A. Rouse. 1967. Population changes Smith, D. W., R. O. Peterson, and D. B. Houston. 2003. Yellowstone after
within the Gallatin elk herd, 1932-65. Journal of Wildlife Manage- wolves. BioScience 53:330–340.
ment 31:304–316. Smith, D. W., D. R. Stahler, D. S. Guernsey. 2006. Yellowstone wolf pro-
Peterson, R. O., J. A. Vucetich, R. E. Page, and A. Chouinard. 2003. ject annual report. Yellowstone Center for Resources, YRC-2006-04.
Temporal and spatial dynamics of predator-prey dynamics. Alces Yellowstone National Park, Mammoth Hot Springs, Wyoming,
39:215–232. Montana.
Ratti, J. T., M. Weinstein, J. M. Scott, P. A. Wiseman, A. Gillesberg, Soulé, M. E., J. E. Estes, B. Miller, and D. L. Honnold. 2005. Strongly in-
C. A. Miller, M. M. Szepanski, and L. K. Svancara. 2004. Feasibility teracting species: conservation, policy, management, and ethics. Bio-
of wolf reintroduction to Olympic Peninsula, Washington. North- Science 55:168–176.
west Science 78:1–76. UDWR (Utah Division of Wildlife Resources). 2005. Utah wolf manage-
Ripple, W. J., and R. L. Beschta. 2003. Wolf reintroduction, predation ment plan. Publication #05-17. UDWR, Salt Lake City, Utah.
risk, and cottonwood recovery in Yellowstone National Park. Forest Vucetich, J. A., D. W. Smith, and D. R. Stahler. 2005. Influence of har-
Ecology and Management 184:299–313. vest, climate, and wolf predation on Yellowstone elk, 1961-2004.
Ripple, W. J., and R. L. Beschta. 2004a. Wolves and the ecology of fear: Oikos 111:259–270.
can predation risk structure ecosystems? BioScience 54:755–766. Wagner, F. H., W. L. Hamilton, and R. B. Keigley. 2006. Yellowstone’s
Ripple, W. J., and R. L. Beschta. 2004b. Wolves, elk, willows, and trophic destabilized ecosystem: elk effects, science, and policy conflicts.
cascades in the upper Gallatin Range of Southwestern Montana, Oxford University Press, Oxford, United Kingdom.
USA. Forest Ecology and Management 200:161–181. White, P. J., and R. A. Garrott. 2005a. Northern Yellowstone elk after
Ripple, W. J., and R. L. Beschta. 2005. Linking wolves and plants: Aldo wolf restoration. Wildlife Society 33:942–955.
Leopold on trophic cascades. BioScience 55:613–621. White, P. J., and R. A. Garrott. 2005b. Yellowstone’s ungulates after
Ripple, W. J., and R. L. Beschta. 2006. Linking wolves to willows via wolves—expectations, realizations, and predictions. Biological
risk-sensitive foraging by ungulates in the northern Yellowstone Conservation 125:141–152.
ecosystem. Forest Ecology and Management 230:96–106. YNP (Yellowstone National Park). 1997. Yellowstone’s northern range:
Ripple, W. J., and R. L. Beschta. 2007. Restoring Yellowstone’s Aspen complexity and change in a wildland ecosystem. National Park
with wolves. Biological Conservation 138:514–519. Service, Mammoth Hot Springs, Wyoming, Montana.

MAY 2010 Restoration Ecology 389