Factors Influencing the Effectiveness of Wildlife Underpasses in Banff National Park, Alberta, Canada

Abstract

Wildlife crossing structures are intended to increase permeability and habitat connectivity across roads. Few studies, however, have assessed the effectiveness of these mitigation measures in a multispecies or community level context. We used a null model to test whether wildlife crossing structures serve large mam-mal species equally or whether such structures limit habitat connectivity across roads in species-specific ways. We also modeled species responses to 14 variables related to underpass structure, landscape features, and hu-man activity. Species performance ratios (observed crossing frequency to expected crossing frequency) were evaluated for four large carnivore and three ungulate species in 11 underpass structures in Banff National Park, Alberta, Canada. Observed crossing frequencies were collected in 35 months of underpass monitoring. Expected frequencies were developed from three independent models: radio telemetry, pellet counts, and hab-itat-suitability indices. The null model showed that species responded to underpasses differently. In the pres-ence of human activity carnivores were less likely to use underpasses than were ungulates. Apart from hu-man activity, carnivore performance ratios were better correlated to landscape variables, and ungulate performance ratios were better correlated to structural variables. We suggest that future underpasses de-signed around topography, habitat quality, and location will be minimally successful if human activity is not managed.

Full text

47
Conservation Biology, Pages 47–56
Volume 14, No. 1, February 2000
Factors Influencing the Effectiveness of Wildlife
Underpasses in Banff National Park, Alberta, Canada
ANTHONY P. CLEVENGER*§ AND NIGEL WALTHO‡
*Faculty of Environmental Design, University of Calgary, Calgary, Alberta T2N 1N4, Canada and Department of
Forestry, Wildlife and Fisheries, The University of Tennessee, Knoxville, Tennessee 37901–1071, U.S.A.
‡Faculty of Environmental Studies, York University, North York, Ontario, M3J 1P3, Canada
Abstract:
Wildlife crossing structures are intended to increase permeability and habitat connectivity across
roads. Few studies, however, have assessed the effectiveness of these mitigation measures in a multispecies or
community level context. We used a null model to test whether wildlife crossing structures serve large mam-
mal species equally or whether such structures limit habitat connectivity across roads in species-specific ways.
We also modeled species responses to 14 variables related to underpass structure, landscape features, and hu-
man activity. Species performance ratios (observed crossing frequency to expected crossing frequency) were
evaluated for four large carnivore and three ungulate species in 11 underpass structures in Banff National
Park, Alberta, Canada. Observed crossing frequencies were collected in 35 months of underpass monitoring.
Expected frequencies were developed from three independent models: radio telemetry, pellet counts, and hab-
itat-suitability indices. The null model showed that species responded to underpasses differently. In the pres-
ence of human activity carnivores were less likely to use underpasses than were ungulates. Apart from hu-
man activity, carnivore performance ratios were better correlated to landscape variables, and ungulate
performance ratios were better correlated to structural variables. We suggest that future underpasses de-
signed around topography, habitat quality, and location will be minimally successful if human activity is not
managed.
Factores que Influencían la Efectivadad de Pasadizos para vida Silvestre en el Parque Nacional Banff, Alberta,
Canada
Resumen:
Las estructuras diseñadas para el cruce de vida silvestre tienen la intención de incrementar la
permeabilidad y conectividad del hábitat a lo largo de las carreteras. Sin embargo, pocos estudios han evalu-
ado la eficacia de estas medidas de mitigación en un contexto multi-especie o de comunidad. Utilizamos un
modelo nulo para evaluar si las estructuras para el cruce de vida silvestre sirven de igual manera a las espe-
cies de mamíferos grandes, o si estas estructuras limitan la conectividad del hábitat a lo largo de carreteras
de manera especie-específica. También modelamos las respuestas de las especies a 14 variables relacionadas
con la estructura de los pasadizos, las características del paisaje y la actividad humana. Se evaluaron tasas
de éxito por especie ( frecuencia de cruces observados/frecuencia de cruces esperados) para cuatro
carnívoros grandes y tres especies de ungulados en 11 estructuras de pasadizos en el parque nacional Banff,
Alberta, Canada. Las observaciones de frecuencias de cruce esperadas se obtuvieron a partir de tres modelos
independientes: radio telemetría, conteo de heces e índices de hábitat adeduado. El modelo nulo mostró que
las especies responden de manera diferente a los pasadizos. En presencia de actividades humanas fue menos
probable que los carnívoros utilizaran los pasadizos en comparación con los ungulados. Aparte de la activ-
idad humana, las tasas de éxito para los carnívoros estuvieron mejor correlacionadas con variables estruc-
turales. Sugerimos que los pasadizos diseñados en el futuro en base a la topografia, la calidad del hábitat y
la ubicación, tendrán un mínimo éxito si la actividad humana no es manejada.
§
Current address: Parks Canada, Box 900, Banff, Alberta T0L 0C0, CANADA, email tony clevenger@pch.gc.ca
Paper submitted February 8, 1999; revised manuscript accepted July 10, 1999.

48
Wildlife Underpasses in Banff National Park Clevenger & Waltho
Conservation Biology
Volume 14, No. 1, February 2000
Introduction
The effect of heavily used roads on mammal populations
has been the focus of many studies during recent years
(Oxley et al. 1974; Mierau & Favara 1975; Mader 1984;
Bennett 1991; Evink et al. 1996). These studies show
that roads affect mammal populations in numerous
ways, from habitat loss and habitat alienation (i.e., sen-
sory disturbance) to physical barriers and road mortality
(Adams & Geis 1983; Mansergh & Scotts 1989; Van der
Zee et al. 1992; Brandenburg 1996). Among these ef-
fects, habitat fragmentation and physical barriers pose
what many conservation ecologists consider the greatest
obstruction to maintaining species diversity and ecologi-
cal integrity (Wilcox & Murphy 1985; Saunders & Hobbs
1991; Dale et al. 1994; Forman & Alexander 1998).
Attempts to increase habitat connectivity and barrier
permeability across road structures can be found in
some road construction and upgrade projects. Wildlife
overpasses and underpasses, for example, first con-
structed in the 1970s, are used as mitigation tools in
many parts of the world today (Reed et al. 1975; Hunt et
al. 1987; Romin & Bissonette 1996; Keller & Pfister
1997). Nonetheless, few studies have examined the effi-
cacy of these mitigation structures (Romin & Bissonette
1996). Furthermore, the few that have been conducted
are limited in their extent to single-species analyses
(Reed et al. 1975; Ballon 1985; Schall et al. 1985; Singer
& Doherty 1985; Woods 1990; Carsignol 1993; but see
Foster & Humphrey 1995). No study has considered the
usefulness of wildlife overpasses and underpasses at
multispecies scales that encompass the large mammals.
Today, highway planners and land managers can ill af-
ford the naïve luxury of single-species mitigation struc-
tures. Species do not function in isolation but are com-
ponents of ecological systems that inherently fall into
the category of organized complexity (Allen & Starr
1982; O’Neill et al. 1986). In an organized, complex sys-
tem, species are dynamically linked to other species on
multiple spatial and temporal scales (Kolasa & Pickett
1989; Pickett et al. 1989, 1997; Waltho & Kolasa 1996;
Fiedler & Kareiva 1998). Therefore, any single-species
mitigation structure is likely to have cascading effects,
some positive and some negative, on nontarget species
also. If a mitigation structure is to succeed, a multispe-
cies approach is needed to evaluate the efficacy of such
mitigation on nontarget species as well.
We evaluated whether underpass structures in Banff
National Park, Alberta, Canada, serve all species (i.e., large
mammals) equally, or whether such structures limit habi-
tat connectivity across roads in species-specific ways. Fur-
thermore, we aimed to determine which of 14 underpass
variables species responded to most, with the anticipa-
tion that once these variables were identified the design
of future mitigation measures could be improved.
Methods
We collected data along the Trans-Canada highway
(TCH) in Banff National Park (BNP), Alberta, Canada
(Fig. 1). The Trans-Canada highway in BNP runs along
the floor of the Bow Valley (2–5 km wide), sharing the
valley bottom with the Bow River, the township of Banff
(population 9000), several high-volume two-lane high-
ways, numerous secondary roads, and the Canadian Pa-
cific Railway. The TCH is the major transportation corri-
dor through the park (park length, 75 km), carrying an
estimated 5 million visitors to the park per year, with an
additional 5 million users en route between Calgary and
Vancouver (Parks Canada Highway Services, unpub-
lished data). The first 45 km of the TCH from the eastern
park boundary (phase 1, 2, and 3A) is four lanes and bor-
dered on both sides by a wildlife exclusion fence 2.4 m
high (phase 1 completed in 1986, phase 2 in 1988, and
phase 3A late 1997). The remaining 30 km to the west-
ern park boundary (Alberta–British Columbia border,
phase 3B) is two lanes and unfenced. Plans exist to up-
grade phase 3B to four lanes with fencing within the
next 5–10 years.
The fenced portion constitutes an effective barrier to
the movement of large mammals. To mitigate this bar-
rier effect, highway engineers constructed 22 wildlife
underpasses and two wildlife overpasses. The effective-
ness of such structures to facilitate large mammal move-
ments, however, is unknown. Because no two under-
passes are similar in all structural and ecological aspects,
we propose that species (i.e., large mammals) select un-
derpasses that best correlate with their ecological needs
and behavior. Attributes that best characterize high-use
underpasses can then be integrated into new designs for
an eventual phase 3B widening process. We tested this
premise at three scales of ecological resolution: species,
species groups, and large mammals. These scales were
used because (1) we anticipate that the explanatory
power of each attribute is dependent, at least in part, on
the ecological resolution used (Rahel et al. 1984; Rahel
1990; Collins & Glenn 1991) and (2) the information
needs of land managers and transportation planners
with respect to mitigation structures can best be met by
a variable scale approach. We chose only phase 1 and
phase 2 underpasses for this study, however, because
the recent completion of phase 3A mitigation structures
does not permit sufficient time for wildlife habituation to
occur at such landscape scales (A.P.C., unpublished data).
Wildlife Underpasses
We monitored 11 wildlife underpasses (Fig. 1): 9 cement
open-span underpasses and 2 metal culverts. We charac-
terized each underpass with 14 variables encompassing
attributes of structural, landscape, and human activity

Conservation Biology
Volume 14, No. 1, February 2000
Clevenger & Waltho Wildlife Underpasses in Banff National Park
49
F
igure 1. Location of study area and 11 wildlife underpasses along the Trans-Canada Highway (TCH) in Banff Na-
tional Park, Alberta. Shaded sections of the TCH refer to phases: lightly shaded, phases 1 and 2; darkly shaded, phase 3A.
(Table 1). Structural variables included underpass width,
height, length (including median), openness (width
⫻
height/length) (Reed & Ward 1985); and noise level
(mean of
A
-weighted decibel readings taken at the center
point within the underpass and 5 m from each end).
Landscape variables included distance to nearest forest
cover, Canadian Pacific Railway, townsite, closest major
drainage, and eastern-most park entrance (hereafter re-
ferred to as east gate). Human activity variables included
types of human use in the underpasses characterized by
counts of people on foot, bike, and horseback, and a hu-
man-use index calculated from the mean monthly counts
of the three former variables combined.
Observed Crossing Frequencies
We measured wildlife use for the 11 underpasses using
methods described by Bider (1968). Specifically, tracking
sections (2
⫻
4 m) were set at both ends of each underpass
to record evidence of underpass use. Tracking material con-
sisted of a dry, loamy mix of sand, silt and clay 3–4 cm deep.
At intervals of 3–4 days, we visited each underpass and clas-
sified the tracking medium as adequate or inadequate, de-
pending on our ability to read tracks clearly. We recorded
species presence (wolves [
Canis lupus
], cougars [
Puma
concolor
], black bears [
Ursus americanus
], grizzly bears
[
U. arctos
], deer [
Odocoileus
sp.], elk [
Cervus elaphus
],
and moose [
Alces alces
]), species abundance, and human
activity at each tracking section during each underpass visit.
Through-passages were recorded for individuals if tracks in
the same direction were present on both tracking sections.
Tracking sections were then raked smooth in preparation
for the next visit. Data were collected in this manner for
two continuous monitoring periods: 1 January 1995–31
March 1996 (15 months) and 1 November 1996–30 June
1998 (20 months). Of 3311 underpass monitoring visits, 59
(1.8%) were classified as inadequate for data collection.
Expected Crossing Frequencies
If the 11 underpasses occur in a homogeneous-habitat
landscape that includes random distribution of species

50
Wildlife Underpasses in Banff National Park Clevenger & Waltho
Conservation Biology
Volume 14, No. 1, February 2000
abundances, then the following assumptions may apply:
(1) the 11 underpasses serve the same population of in-
dividuals and (2) each individual is aware of all 11 under-
passes and can choose between underpasses based on
underpass attributes alone. The Banff Bow Valley is a
highly heterogeneous landscape, that is, lakes, mountain
barriers, and narrow corridors (for example) restrict un-
derpass accessibility on multiple spatiotemporal scales.
If habitat fragmentation is perceived as extreme, then
we may assume that each underpass serves its own
unique subpopulation. If this were true, then differ-
ences in observed crossings frequencies between under-
passes would reflect differences in subpopulation sizes
alone and not attributes of the underpasses themselves.
Although these two sets of assumptions represent end-
points along a continuum of possible interactions, the
relative extent that species interact with the habitat
landscape and distribution of underpasses is unknown.
It is therefore necessary to examine observed crossing
frequencies in the context of expected crossing frequen-
cies (i.e., performance ratios).
Expected crossing frequencies were obtained from
three independent data sets that included radiotelemetry
location data, relative-abundance pellet transects, and
habitat-suitability indices. Because it remains unclear
what proportion of individuals from these data sets uses
the underpasses directly, we defined our expected cross-
ing frequencies as equal to the abundance data found at
radii 1, 2, and 3 km from the center of each underpass.
Specifically, we used (1) radiotelemetry location data for
black bears (
n
⫽
255 locations), grizzly bears (
n
⫽
221 lo-
cations), wolves (
n
⫽
2314 locations), and elk (
n
⫽
1434
locations; Parks Canada, unpublished data); (2) relative-
abundance pellet transects for deer (
n
⫽
1579 pellet
sites), elk (
n
⫽
26,614 pellet sites), moose (
n
⫽
43 pellet
sites), and wolves (
n
⫽
30 sites containing scat; Parks
Canada, unpublished data); and (3) habitat suitability indi-
ces for black bears, cougars, wolves, deer, elk, and moose
(Holroyd & Van Tighem 1983; Agriculture Canada 1989;
Kansas & Raines 1990).
Analyses
We derived species-performance ratios for each of the
three independent data sets by dividing observed cross-
ing frequencies by expected crossing frequencies. Per-
formance ratios were designed such that the higher the
ratio, the more effectively the underpass appeared to fa-
cilitate species crossings.
We examined the premise that wildlife crossing struc-
tures serve species equally by testing the null hypothesis
that performance ratios do not differ between species
(paired
t
test with Bonferroni adjusted probability val-
ues; SYSTAT 1998). We tested the null hypothesis for
each of the three performance models—radiotelemetry,
habitat suitability indices, and log-transformed pellet
counts—partly because no one model includes the com-
plete species composition.
In the event that we rejected the null hypothesis, we
proceeded with three steps to determine which of the
14 underpass attributes were most closely associated
with species performance ratios. First, we standardized
all performance ratios to a mean of zero and a standard
deviation of one to remove absolute differences be-
tween the three models.
Second, we used a family of simple curvilinear and
polynomial regression curves to optimize the fit be-
tween species-performance ratios and each underpass
Table 1. Attributes of 11 wildlife underpasses used in analysis of factors influencing wildlife in Banff National Park, Alberta.
Underpass
Underpass attribute 1 234567891011
Structural
width (m) 9.8 13.4 4.2 9.8 9.5 14.9 10.0 9.8 10.3 9.0 7.0
height (m) 2.8 2.5 3.5 2.9 2.9 3.2 3.0 2.7 2.8 2.9 4.0
length (m) 63.0 83.2 96.1 40.0 39.7 38.0 27.1 27.2 25.6 40.1 56.0
openness 0.43 0.4 0.15 0.71 0.69 1.25 1.1 0.97 1.12 0.65 0.5
noise level
a
68.1 70.5 64.1 66.8 66.0 63.8 64.3 67.4 67.4 67.1 64.1
Landscape (distance to)
east gate (km) 0.0 2.1 3.5 5.8 10.5 11.5 12.0 14.4 17.0 18.8 38.8
forest cover (m) 22.3 63.3 11.9 15.2 47.3 16.1 35.9 23.3 27.5 23.9 35.4
nearest drainage (km) 1.0 0.0 0.1 0.4 0.6 0.0 0.6 1.2 0.4 0.2 0.3
Canadian Pacific Railway track (km) 0.5 0.75 0.8 0.02 0.02 0.02 0.25 1.2 0.4 0.75 0.75
nearest town (km) 1.6 3.5 5.5 6.0 1.5 0.5 0.2 1.7 5.2 7.2 0.8
Human activity
human-use index
b
0.4 1.9 1.8 0.6 5.3 5.3 15.2 3.2 11.4 0.6 0.5
bike 0 5 6 21 189 8 462 19 595 1 0
horseback 6 3 6 5 42 138 186 12 58 10 10
foot 7 45 14 20 34 77 129 80 241 10 29
a
Mean of A-weighted decibel readings taken at the center point within the underpass and 5 m from each end.
b
Calculated from the mean monthly counts of people on foot, bike, and horseback.

Conservation Biology
Volume 14, No. 1, February 2000
Clevenger & Waltho Wildlife Underpasses in Banff National Park
51
attribute ( Jandel Scientific 1994). We used the following
criteria to choose the optimal equation for each regres-
sion analysis. (1) the regression model had to be statisti-
cally significant (at
p
⬍
0.05). (2) The beta coefficient
for the highest ordered term had to be statistically signif-
icant. (3) Once an equation met the above criteria, we
compared its
F
statistic with the
F
statistic for the next
equation that also met these criteria but had one less or-
dered term. We chose the model with the higher
F
sta-
tistic and (4) iterated the above process for equations
with consecutively fewer terms. (5) If no curvilinear or
polynomial equation was accepted, we chose the simple
linear regression model (equation 41; Appendix 1) to de-
scribe the relationship, assuming that it had not already
been chosen through the iterative process. (6) If these
criteria failed to produce a significant regression model
for species per se and underpass attribute per se, we de-
leted the underpass attribute as being a significant factor
influencing the species-performance ratio.
Third, for each species we ranked the regression mod-
els thus obtained according to the absolute value of each
model’s coefficient of determination. This three-step
process allowed for the identification and ordering of
underpass attributes (in order of importance) associated
with each species performance ratio, but it failed to sep-
arate ecologically significant attributes from those that
appeared significant but were statistical artifacts of the
underpasses themselves.
The three-step process was repeated for each of the
three scales of ecological resolution. For species groups,
however, it was first necessary to identify group types
according to similarities in species performance ratios as
compared to some arbitrary definition. We used princi-
pal component analysis (PCA) to identify these species
groups. Because none of the performance models con-
tains a full species list, it was necessary to include all
species performance ratios from each of the models into
the single PCA.
Results
From 1 January 1995 to 30 June 1998 (excluding 1 April
to 31 October 1996) 14,592 large-mammal underpass vis-
its were recorded. Ungulates were 78% of this total, carni-
vores 5%, and human-related activities 17% (Table 2). Indi-
vidual underpasses ranged from 373 visits to 2548 visits.
Specific to wildlife, elk were the most frequently ob-
served species (
n
⫽
8959, 74% of all wildlife), followed
by deer (
n
⫽
2411, 20%), and then wolves (
n
⫽
311,
2.5%). The through-passage rate for wildlife species was
high (mean 98%, SD
⫽
1.9).
For each underpass, species-performance ratios signif-
icantly differed between species (paired
t
test with Bon-
ferroni adjusted probability;
p
⬍
0.001). We therefore
rejected the null hypothesis and focused instead on
identifying the underpass attributes that most likely in-
fluenced a species’s underpass use.
For individual species, the rank order of significant at-
tributes was not significantly different between perfor-
mance models (paired
t
test, all within-species compari-
sons not significant at
p
⬍
0.05). We therefore provide
mean rank scores only (Table 3). The rank order of sig-
nificant attributes, however, does differ between spe-
cies (paired
t
test, Bonferroni adjusted probability val-
ues;
p
⬍
0.05). For example, we found that underpass
distance from the east gate (positive correlation) was
the most significant underpass attribute affecting black
bear performance ratios, whereas underpass length
(negative correlation) was the most significant attribute
affecting elk performance ratios (Table 3).
At the second scale of ecological resolution, species
groups, we used PCA to identify two group types (Fac-
tor 1, Fig. 2). The two groups were readily identifiable as
large predators/omnivores (hereafter referred to as car-
nivores) and ungulates. For carnivores the most signifi-
cant underpass attribute influencing the group’s perfor-
mance was distance to townsite (positively correlated),
followed by human activities such as hiking (negatively
correlated), human use index (negatively correlated),
and horseback riding (negatively correlated). Landscape
and structural variables were the least significant at-
tributes influencing the group’s performance ratio (i.e.,
distance to nearest drainage, negatively correlated; un-
derpass openness, negatively correlated; Table 4).
In contrast, we found that the most significant under-
pass attributes influencing ungulates were structural and
landscape factors. Specifically, the rank order was 1, un-
derpass openness (negatively correlated); 2, noise level
(positively correlated); 3, underpass width (negatively
correlated), and 5, distance to nearest drainage. Human
activity attributes, although significant, were ranked
lower: 4, horseback riding (negatively correlated), and
6, hiking (negatively correlated; Table 4).
At the third scale of ecological resolution, large mam-
mals (i.e., all species together), we found that the most
significant underpass attribute influencing the commu-
nity’s performance ratio was structural openness (nega-
tively correlated; Table 4). Distance to townsites was the
second most significant attribute (positive correlation),
followed by human activity (human-use index, horseback
riding, hiking, and biking, all negatively correlated).
Discussion
There were no significant differences in the rank order of
the 14 underpass attributes between the three performance
ratio models (radiotelemetry, pellet count, and habitat suit-
ability indices). This suggests that although the subpopula-
tion that each underpass serves is unknown our confidence
in using performance ratios as a means to standardize differ-

52
Wildlife Underpasses in Banff National Park Clevenger & Waltho
Conservation Biology
Volume 14, No. 1, February 2000
ences in species abundance between underpasses is high.
More importantly, however, these results permit us to test
the null hypothesis independently of the actual grain in
which species interact with the habitat template.
Our results suggest that underpass attributes differen-
tially influence species performance ratios. Depending
on the ecological resolution (i.e., species, species
groups, large mammals), however, different underpass
attributes were interpreted as dominant. One common
thread at all resolutions was that human influence—
whether it was distance to townsite or human activity
within an underpass—consistently ranked high as a sig-
nificant factor affecting species-performance ratios. At
the species level, for example, results from six of the
seven species ranked at least one human attribute as the
first or second most important attribute influencing
the species-performance ratio. At the group level, carni-
vores showed a positive correlation between underpass
performance ratios and distance from town and a nega-
tive correlation to human activity. The inverse relation
between the two human
-
related attributes occurred be-
cause the townsites served as sources of human popula-
tions from which human activity originates. The closer
an underpass was to the town of Banff or Canmore, the
greater the human use activity observed (Mattson et al.
1987; Kasworm & Manley 1990; McCutchen 1990;
Jalkotzy & Ross 1993; Gibeau et al. 1996; Paquet et al.
1996; but see Rodriguez et al. 1997).
Ungulates, however, failed to respond to human activ-
ity in the same manner. Although significant negative
correlations in performance ratios were observed, the
relative importance of human activity ranked below that
of structural attributes. Elk habituation to human pres-
ence close to town may, at least in part, have masked
the performance ratios of unhabituated elk farther from
town (Woods et al. 1996). At the community level, the
most important attribute influencing species perfor-
mance ratios was structural openness. The second most
important attribute, however, was distance to the town-
sites (positive correlation).
These results lend support to the Banff National Park
management plan, which emphasizes stricter limits on
human development and more effective methods of
managing and limiting human use within the park (Parks
Canada 1997). The BNP management plan also recom-
mends improving the effectiveness of phase 1 and 2 un-
derpasses by “retrofitting.” In this context we suggest
that in such a multispecies system the most efficient
Table 2. Observed use of wildlife underpasses by carnivores and ungulates in Banff National Park, Alberta, 1995–1998.
Underpasses
Species 1 2 3 4 5 6 7 8 9 10 11
Black bear 10 20 43 37 13 8 0 4 8 34 16
Grizzly bear000 2 00 0 0050
Cougar 5 29 3 30 7 0 4 4 20 15 0
Wolf 1 7 3 28 3 5 1 5 77 146 35
Deer 554 42 294 253 215 21 61 338 288 291 54
Elk 825 201 331 1199 1062 467 1576 1522 821 683 272
Moose 1 0 1 0 0 0 0 0000
Table 3. Species level rank ordering of mean coefficient of determinations and their slope for models explaining underpass interactions in
Banff National Park, Alberta.
Underpass attributes Black bear Grizzly bear Cougar Wolf Deer Elk Moose
Width 8
⫺
4
⫺
3
⫺
5
⫺
Height 3
⫺
3
⫹
10
⫹
Length 7
⫹
1
⫺
4
⫹
Openness 4
⫺
5
⫺
4
⫹
1
⫺
Noise level 12
⫹
1
⫹
3
⫹
8
⫹
Distance to
east gate 1
⫺
2
⫹
3
⫺
forest cover 11
⫺
3
⫺
4
⫹
6
⫺
11
⫺
6
⫺
nearest drainage 9
⫺
7
⫺
2
⫹
2
⫹
Canadian Pacific Railway
track 4
⫺
5
⫹
8
⫹
nearest town 3
⫹
1
⫹
2
⫹
1
⫹12 ⫹
human activity
human-use index 6 ⫺2 ⫺6 ⫺5 ⫺8 ⫺
bike 10 ⫺4 ⫺6 ⫺7 ⫺
horseback 5 ⫺1 ⫺7 ⫺2 ⫺
foot 2 ⫺5 ⫺8 ⫺7 ⫺9 ⫺9 ⫺

Conservation Biology
Volume 14, No. 1, February 2000
Clevenger & Waltho Wildlife Underpasses in Banff National Park 53
(and probably economic) approach to retrofitting is to
manage human activity near each underpass. Specifi-
cally, we recommend that foot trails be relocated and
human use of underpasses be restricted. Continual mon-
itoring of wildlife passage frequencies at these struc-
tures will permit Parks Canada to evaluate how this man-
agement strategy may translate into greater permeability
of the Trans-Canada Highway and habitat connectivity
for all wildlife populations in the Bow Valley.
Landscape variables other than distance to town may
also be important attributes influencing species-perfor-
mance ratios. Carnivores had a greater tendency to use
underpasses close to drainages systems, for example,
whereas ungulates tended to avoid them. Drainage sys-
tems are known travel routes for wildlife, particularly in
narrow glacial valleys such as Banff’s Bow Valley. The in-
verse relationship between carnivores and ungulates
with respect to drainages may reflect processes such as
predator-prey interactions rather than any direct effect
of landscape attributes on underpass use. Recent studies
have shown that predators can have important effects
on the community structure of prey species (Lima & Dill
1990; Dickman 1992; Jedrzejewski et al. 1993). For ex-
ample, deer are known to keep to the periphery of wolf
territories (Hoskinson & Mech 1976; Mech 1977) and re-
duce their feeding effort when exposed to odors of
wolves and other predator species (Muller-Schwarze
1972; Sullivan et al. 1985). Furthermore, there is some
evidence that the presence of badgers (Meles meles) can
disrupt their prey species (hedgehogs [Erinacceus euro-
paeus]) use of tunnels under roads in England (C. Don-
caster, unpublished data).
The results from our analyses also suggest that struc-
tural attributes were significant in species-performance
ratios, especially for ungulates. Ungulates preferred un-
derpass structures with a low openness ratio, narrow
width, and long tunnel dimensions. We doubt, however,
that such species prefer constricted underpasses over
larger and more open underpasses. Previous studies
have shown that ungulates were reluctant to use under-
passes ⬍7 m wide or ⬍2.4 m high (Reed et al. 1975;
Yanes et al. 1995; Rosell et al. 1997). Therefore, in a se-
ries of post-hoc regression analyses, we found that open-
ness was significantly correlated to length, noise, and
distance to town (linear regression, p ⬍ 0.05). These
post-hoc tests suggested that the importance of these
structural attributes may be statistical artifacts.
Although there is limited information on the suitability
of underpass design for large carnivores (but see Rod-
riguez et al. 1997), it is understood underpasses that are
long and low in clearance inhibit use by carnivores
(Hunt et al. 1987; Beier & Loe 1992; Foster & Humphrey
1995). Results from our analyses agree in part with this
expectation because wolf performance ratios were posi-
tively correlated with underpass height; for other carni-
vore species, however, attributes of underpass structure
contributed little.
It is possible that the overall weakness of structural at-
tributes in explaining species performance ratios may be
Figure 2. Principal component analysis of models of
observed and expected underpass use for wildlife in
Banff National Park, Alberta. Models were developed
from observation data (OBS), radiotelemetry location
data (TEL), relative-abundance pellet-transect data
(PEL), and habitat suitability index data (HSI) for
black bears (BB), grizzly bears (GB), cougars (C),
wolves (W), deer (D), elk (E), and moose (M). Two dis-
crete groups were identified along factor 1: large pred-
ators/omnivores (
䊉
) and herbivores (
䊊
).
Table 4. Rank ordering of mean coefficients of determination and
their slope for models explaining underpass interactions at the level
of species groups and large mammals in Banff National Park,
Alberta.
Underpass
attributes Carnivores Ungulates
Large
mammals
Width 3 ⫺6 ⫺
Height 10 ⫺
Length 8 ⫹11 ⫹
Openness 5⫺1 ⫺1 ⫺
Noise level 7⫹2 ⫹8 ⫹
Distance to
east gate 10 ⫺13 ⫹
forest cover 7 ⫺12 ⫺
nearest drainage 6 ⫺5 ⫹
Canadian Pacific
Railway track 12 ⫹9 ⫹
nearest town 1⫹13 ⫹2 ⫹
Human activity
human-use index 3 ⫺9 ⫺3 ⫺
bike 8 ⫺11 ⫺7 ⫺
horseback 4 ⫺4 ⫺4 ⫺
foot 2 ⫺6 ⫺5⫺

54 Wildlife Underpasses in Banff National Park Clevenger & Waltho
Conservation Biology
Volume 14, No. 1, February 2000
due to each species’s individual familiarization with the
12-year-old underpasses. Individuals require time to
adapt to underpass structures (Reed et al. 1975; Waters
1988; Bunyan 1990; Land & Lotz 1996; A.P.C., unpub-
lished data); once adaptation has occurred, the dynam-
ics of human activity and attributes of landscape hetero-
geneity may play a larger role in determining species-
performance ratios than the structural attributes them-
selves (Gibeau & Herrero 1998).
The underpass attributes varied in importance be-
tween both species and ecological resolutions. The mul-
tiscale approach we used demonstrates that the informa-
tional needs of a state transportation planner responsible
for site-specific mitigation for deer (Reed et al. 1975;
Romin & Bissonette 1996) will likely be different from
those of a land manager in BNP mandated to maintain
ecosystem integrity of a 650,000-ha national park. Inde-
pendent of the ecological resolution used, however, spe-
cies-performance ratios were consistently negatively cor-
related with some measure of human activity. Therefore,
the best designed and landscaped underpasses may be in-
effective if human activity is not controlled.
Acknowledgments
J. Mamalis, B. Bertch, A. Whitehead, Z. Callaway, K.
Wells, P. Smillie, C. St-Pierre, and M. Brumfit assisted in
the data collection. We thank Parks Canada Highway
Services Centre (PCHSC) and Banff National Park, World
Wildlife Fund-Canada, and the Agricultural Research
Foundation at Oregon State University for providing
funding for the research. T. McGuire (PCHSC) was in-
strumental in providing necessary funds for the project.
T. Hurd and C. White (Banff National Park) helped se-
cure administrative and logistical support. D. Zell pro-
vided support for the geographic information system.
The following researchers generously allowed us to use
their data for this analysis: M. Hebblewhite, C. Cal-
laghan, P. Paquet, M. Gibeau, S. Herrero, and T. Hurd.
We thank M. Hourdequin, E. Main, G. Meffe, and two
anonymous reviewers for critically reading earlier ver-
sions of this manuscript.
Literature Cited
Adams, L. W., and A. D. Geis. 1983. Effects of roads on small mammals.
Journal of Applied Ecology 20:403–415.
Agriculture Canada. 1989. Ecosites of the Lower Bow Valley Banff Na-
tional Park East Gate-Castle Junction. Map Reproduction Centre, De-
partment of Energy, Mines and Resources, Ottawa, Ontario, Canada.
Allen T. F. H., and T. B. Starr, editors. 1982. Hierarchy: perspectives for
ecological complexity. University of Chicago Press, Chicago.
Ballon, P. 1985. Premieres observations sur l’efficacie des passages a
gibier sur l’autoroue A36. Pages 311–316 in Routes et faune sau-
vage. Service d’Etudes Techniques de Routes et Autoroutes, Bag-
neaux, France.
Beier, P., and S. Loe. 1992. A checklist for evaluating impacts to wild-
life movement corridors. Wildlife Society Bulletin 20:434–440.
Bennett, A. F. 1991. Roads, roadsides and wildlife conservation: a re-
view. Pages 99–118 in D.A. Saunders and R.J. Hobbs, editors. Na-
ture conservation 2: the role of corridors. Surrey Beatty & Sons,
Chipping Norton, New South Wales.
Bider, J. R. 1968. Animal activity in uncontrolled terrestrial communi-
ties as determined by a sand transect technique. Ecological Mono-
graphs 38:274–291.
Brandenburg, D. M. 1996. Effects of roads on behavior and survival of
black bears in coastal North Carolina. M.S. thesis. University of Ten-
nessee, Knoxville.
Bunyan, R. 1990. Monitoring program of wildlife mitigation measures.
Trans-Canada Highway twinning: phase II. Final report. Parks Can-
ada, Banff National Park, Alberta, Canada.
Carsignol, J. 1993. Passages pour la grande faune. Atelier Central de
l’Environnement, Paris.
Collins S. L., and S. M. Glenn. 1991. Importance of spatial and tempo-
ral dynamics in species regional abundance and distribution. Ecol-
ogy 72:654–664.
CTGREF. 1978. Autoroute et grand gibier. Note technique 42. Groupe-
ment technique forestier. Ministere de l’Agriculture, Paris.
Dale, V. H., S. M. Pearson, H. L. Offerman, and R. V. O’Neill. 1994. Re-
lating patterns of land-use change to faunal biodiversity in the Cen-
tral Amazon. Conservation Biology 8:1027–1036.
Dickman, C.R. 1992. Predation and habitat shift in the house mouse,
Mus domesticus. Ecology 73:313–322.
Evink, G., D. Ziegler, P. Garrett, and J. Berry. 1996. Highways and
movement of wildlife: improving habitat connections and wildlife
passageways across highway corridors. Proceedings of the Florida
Department of Transportation/Federal Highway Administration
transportation-related wildlife mortality seminar. Florida Depart-
ment of Transportation, Tallahassee.
Fiedler P., and P. M. Kareiva, editors. 1998. Conservation biology for
the coming decade. Chapman & Hall, New York.
Forman, R. T. T., and L. E. Alexander. 1998. Roads and their major ecolog-
ical effects. Annual Review of Ecology and Systematics 29:207–231.
Foster, M. L., and S. R. Humphrey. 1995. Use of highway underpasses
by Florida panthers and other wildlife. Wildlife Society Bulletin 23:
95–100.
Gibeau, M. G., and S. Herrero. 1998. Roads, rails and grizzly bears in
the Bow River Valley, Alberta. Pages 104–108 in G. Evink, D. Zie-
gler, P. Garrett, and J. Berry, editors. Proceedings of the interna-
tional conference on wildlife ecology and transportation. Publica-
tion FL–ER–69–98. Florida Department of Transportation,
Tallahassee.
Gibeau, M. G., S. Herrero, J. L. Kansas, and B. Benn. 1996. Grizzly bear
population and habitat status in Banff National Park. Report. Banff-
Bow Valley Task Force, Banff, Alberta.
Holroyd, G. L., and K. J. Van Tighem. 1983. Ecological (biophysical)
land classification of Banff and Jasper national parks. Volume 3.
The wildlife inventory. Canadian Wildlife Service, Edmonton, Al-
berta, Canada.
Hoskinson, R. L., and L. D. Mech. 1976. White tailed deer migration and its
role in wolf predation. Journal of Wildlife Management 40:429–441.
Hunt, A., H. J. Dickens, and R. J. Whelan. 1987. Movement of mammals
through tunnels under railway lines. Australian Zoologist 24:89–93.
Jalkotzy, M., and I. Ross. 1993. Cougar responses to human activity at
Sheep River, Alberta. Arc Wildlife Services, Calgary, Alberta, Canada.
Jandel Scientific. 1994. Tablecurve 2D. Version 3 for Win32. Jandel Sci-
entific, San Rafael, California.
Jedrzejewski, W., L. Rychlik, and B. Jedrzejewska. 1993. Responses of
bank voles to odours of seven species of predators: experimental
data and their relevance to natural predator-vole relationships. Oi-
kos 68:251–257.
Kansas, J. L., and R. M. Raines. 1990. Methodologies used to assess the
relative importance of ecological land classification units to black

Conservation Biology
Volume 14, No. 1, February 2000
Clevenger & Waltho Wildlife Underpasses in Banff National Park 55
bears in Banff National Park, Alberta. International Conference on
Bear Research and Management 8:155–160.
Kasworm, W., and T. Manley. 1990. Road and trail influences on griz-
zly and black bears in northwest Montana. International Confer-
ence on Bear Research and Management 8:79–84.
Keller, V., and H. P. Pfister. 1997. Wildlife passages as a means of miti-
gating effects of habitat fragmentation by roads and railway lines.
Pages 70–80 in K. Canters, editor. Habitat fragmentation & infra-
structure. Ministry of Transportation, Public Works & Water Man-
agement, Delft, The Netherlands.
Kolasa J., and S. T. A. Pickett. 1989. Ecological systems and the con-
cept of biological organization. Proceeding of the U.S. National
Academy of Sciences 86:8837–8841.
Land, D., and M. Lotz. 1996. Wildlife crossing designs and use by Flor-
ida panthers and other wildlife in southwest Florida. Pages 323–
328 in G. Evink, D. Ziegler, P. Garrett, and J. Berry, editors. High-
ways and movement of wildlife: improving habitat connections
and wildlife passageways across highway corridors. Florida Depart-
ment of Transportation, Tallahassee.
Lima, S. L., and L. M. Dill. 1990. Behavioral decisions made under the
risk of predation: a review and prospectus. Canadian Journal of Zo-
ology 68:619–640.
Mansergh, I. M., and D. J. Scotts. 1989. Habitat continuity and social or-
ganization of the mountain pygmy–possum restored by tunnel.
Journal of Wildlife Management 53:701–707.
Mattson, D. J., R. R. Knight, and B. M. Blanchard. 1987. The effects of
developments and primary roads on grizzly bear habitat use in Yel-
lowstone National Park, Wyoming. International Conference on
Bear Research and Management 7:259–273.
Mader, H. J. 1984. Animal habitat isolation by roads and agricultural
fields. Biological Conservation 29:81–96.
McCutchen, H. E. 1990. Cryptic behavior of black bears in Rocky
Mountain National Park, Colorado. International Conference on
Bear Research and Management 8:65–72.
Mech, L. D. 1977. Wolf-pack buffer zones as prey reservoirs. Science
198:320–321.
Mierau, G. W., and B. E. Favara. 1975. Lead poisoning in roadside pop-
ulations of deer mice. Environmental Pollution 8:55–64.
Muller-Schwarze, D. 1972. Responses of young black-tailed deer to
predator odours. Journal of Mammalogy 53:393–394.
O’Neill, R. V., D. L. DeAngelis, J. B. Waide, and T. F. H. Allen, editors.
1986. A hierarchical concept of ecosystems. Princeton University
Press, Princeton, New Jersey.
Oxley, D. J., M. B. Fenton, and G. R. Carmody. 1974. The effects of
roads on small mammals. Journal of Applied Ecology 11:51–59.
Paquet, P. C., J. Wierczhowski, and C. Callaghan. 1996. Summary re-
port on the effects of human activity on gray wolves in the Bow
River Valley, Banff National Park, Alberta. Report. Banff-Bow Valley
Task Force, Banff, Alberta, Canada.
Parks Canada. 1997. Banff National Park management plan. Ministry of
Canadian Heritage, Ottawa, Ontario, Canada.
Pickett, S. T. A., J. Kolasa, S. Collins, and J. J. Armesto. 1989. The eco-
logical concept of disturbance and its expression at various hierar-
chical levels. Oikos 54:129–136.
Pickett, S. T. A., R. S. Ostfeld, M. Shackak, and G. E. Likens, editors.
1997. The ecological basis of conservation: heterogeneity, ecosys-
tems, and biodiversity. Chapman & Hall, New York.
Rahel, F. J 1990. The hierarchical nature of community persistence: a
problem of scale. American Naturalist 136:328–344.
Rahel, F. J., J. D. Lyons, and P. A. Cochran. 1984. Stochastic or deter-
ministic regulation of assemblage structure? It may depend on how
the assemblage is defined. American Naturalist 124:583–589.
Reed, D. F., and A. L. Ward. 1985. Efficacity of methods advocated to
reduce deer-vehicle accidents: research and rationale in the USA.
Pages 285–293 in Routes et faune sauvage. Service d’Etudes Tech-
niques de Routes et Autoroutes, Bagneaux, France.
Reed, D. F., T. N. Woodward, and T. M. Pojar. 1975. Behavioral re-
sponse of mule deer to a highway underpass. Journal of Wildlife
Management 39:361–367.
Rodriguez, A., G. Crema, and M. Delibes. 1997. Factors affecting cross-
ing of red foxes and wildcats through non-wildlife passages across
a high–speed railway. Ecography 20:287–294.
Romin, L. A., and J. A. Bissonette. 1996. Deer-vehicle collisions: status
of state monitoring activities and mitigation efforts. Wildlife Society
Bulletin 24:276–283.
Rosell, C., J. Parpal, R. Campeny, S. Jove, A. Pasquina, and J. M. Ve-
lasco. 1997. Mitigation of barrier effect on linear infrastructures on
wildlife. Pages 367–372 in K. Canters, editor. Habitat fragmenta-
tion and infrastructure. Ministry of Transportation, Public Works
and Water Management, Delft, The Netherlands.
Saunders, D. A., and R. J. Hobbs. 1991. Nature conservation 2: the role
of corridors. Surrey Beatty & Sons, Chipping Norton, New South
Wales.
Schall, A., L. D. Humblot, and D. Guilminot. 1985. Premieres donnees
sur la frequentation de passages a faune par la cerf, Autoroute A26,
Haute Marne, Nord-Est de la France. Pages 269–274 in Routes et
faune sauvage. Service d’Etudes Techniques de Routes et Autor-
outes, Bagneaux, France.
Singer, F. J., and J. L. Doherty. 1985. Managing mountain goats at a
highway crossing. Wildlife Society Bulletin 13:469–477.
Sullivan, T. P., L. O. Nordstrom, and D. S. Sullivan. 1985. Use of preda-
tor odors as repellents to reduce feeding damage by herbivores II.
Black-tailed deer (Odocoileus hemionus columbianus). Journal of
Chemical Ecology 11:921–935.
SYSTAT. 1998. Version 8.0 for Windows. Evanston, Illinois.
Van der Zee, F. F., J. Wiertz, C. J. F. Ter Braak, R. C. van Apeldorn, and
J. Vink. 1992. Landscape change as a possible cause of the badger
Meles meles L. decline in The Netherlands. Biological Conservation
61:17–22.
Waltho, N., and J. Kolasa. 1996. Stochastic determinants of assemblage
patterns in coral reef fishes: a quantification by means of two mod-
els. Environmental Biology of Fishes 47:255–267.
Waters, D. 1988. Monitoring program mitigative measures, Trans-Can-
ada Highway twinning: phase I. Final report. Parks Canada, Banff
National Park, Alberta, Canada.
Wilcox, B. A., and D. D. Murphy. 1985. Conservation strategy: the effects
of fragmentation on extinction. American Naturalist 125:879–887.
Woods, J. G. 1990. Effectiveness of fences and underpasses on the
Trans-Canada Highway and their impact on ungulate populations
project. Report. Parks Canada, Banff National Park Warden Service,
Banff, Alberta, Canada.
Woods J. G., L. Cornwell, T. Hurd, R. Kunelius, P. Paquet, and J.
Wierzchowski. 1996. Elk and other ungulates. Pages 149–156 in J.
Green, C. Pacas, L. Cornwell, and S. Bayley, editors. Ecological out-
looks project: a cumulative effects assessment and futures outlook
of the Banff Bow Valley. Department of Canadian Heritage, Ottawa,
Ontario, Canada.
Yanes, M., J. M. Velasco, and F. Suarez. 1995. Permeability of roads and
railways to vertebrates: the importance of culverts. Biological Con-
servation 71:217–222.

56 Wildlife Underpasses in Banff National Park Clevenger & Waltho
Conservation Biology
Volume 14, No. 1, February 2000
Appendix 1
Curvilinear regression equations used to optimize the fit between
species-performance ratios and each underpass attribute.
Number Equation
1lny
⫽
a⫹bx⫹cx2⫹dx3
2y2
⫽
a⫹bx⫹cx2⫹dx3
3y0.5
⫽
a⫹bx⫹cx2⫹dx3
4y-1
⫽
a⫹bx⫹cx2⫹dx3
5y
⫽
a⫹b(lnx)⫺1⫹c(lnx)⫺2⫹d(lnx)⫺3
6y
⫽
a⫹b(x)⫺1⫹c(x)⫺2⫹d(x)⫺3
7y
⫽
a⫹blnx⫹c(lnx)2⫹d(lnx)3
8y
⫽
a⫹b(lnx)2⫹clnx⫹d(lnx)⫺1
9y
⫽
a⫹bx⫹cx2⫹d(x)⫺1
10 y
⫽
a⫹bx⫹cx2⫹dx3
11 lny
⫽
a⫹bx⫹cx2
12 y2
⫽
a⫹bx⫹cx2
13 y0.5
⫽
a⫹bx⫹cx2
14 y
⫽
a⫹b(lnx)⫺1⫹c(lnx)⫺2
15 y
⫽
a⫹b(x)⫺1⫹c(x)⫺2
16 y
⫽
a⫹blnx⫹c(lnx)⫺1
17 y
⫽
a⫹b(lnx)2⫹clnx
18 y
⫽
a⫹bx⫹c(x)⫺1
19 y
⫽
a⫹bx⫹cx2
20 lny
⫽
a⫹bx
21 y
⫽
a⫹be
⫺
x
22 y
⫽
a⫹b(x)⫺2
23 y
⫽
a⫹blnx(x)⫺2
24 y
⫽
a⫹b(x)⫺1.5
25 y
⫽
a⫹b(x)⫺1
26 y
⫽
a⫹blnx(x)⫺1
27 y
⫽
a⫹b(x)⫺0.5
28 y
⫽
a⫹b(lnx)⫺1
29 y
⫽
a⫹blnx
30 y
⫽
a⫹bx0.5
31 y
⫽
a⫹bx(lnx)⫺1
32 y
⫽
a⫹b(lnx)2
33 y
⫽
a⫹bx0.5 lnx
34 y
⫽
a⫹bex
35 y
⫽
a⫹bx3
36 y
⫽
a⫹bx2.5
37 y
⫽
a⫹bx2 lnx
38 y
⫽
a⫹bx2
39 y
⫽
a⫹bx1.5
40 y
⫽
a⫹bxlnx
41 y
⫽
a⫹bx