Biological Conservation 115 (2004) 487–497
www.elsevier.com/locate/biocon
Landscape effects on black-tailed prairie dog colonies
Whitney C. Johnsona,*, Sharon K. Collingeb
a
Department of Environmental, Population, and Organismic Biology, University of Colorado at Boulder,
Boulder, CO 80309-0334, USA
b
Environmental Studies Program, University of Colorado at Boulder, Boulder, CO 80309-0334, USA
Received 20 August 2002; received in revised form 7 February 2003; accepted 19 March 2003
Abstract
Black-tailed prairie dogs (Cynomys ludovicianus) increasingly compete for available habitat with human development in the
Colorado Front Range. Because the effects of increased urbanization on prairie dog colonies are unknown, we studied how landscape context affects prairie dog density in Boulder County, Colorado, USA. We used burrow density as a proxy for prairie dog
density because these variables were correlated at our study sites (r=0.60). Using remotely sensed data and a GIS, we quantified
percent urbanization, road density, and the percentage of other prairie dog colonies in the surrounding landscape at 200, 1000, and
2000 m from the perimeter of 22 prairie dog colonies, and compared burrow density with each landscape variable at each scale. We
also calculated Akaike’s information criterion (AIC) to determine the most parsimonious models predicting burrow density. Ranges
of burrow densities and prairie dog densities in Boulder County were higher than in other studies using similar methodology.
Within Boulder County, burrow density was significantly higher in colonies surrounded by greater density of roads. The degree to
which prairie dog colonies were immediately surrounded by unsuitable habitat, i.e. the ‘‘boundedness’’ of the colony, was negatively
correlated with colony area and positively correlated with burrow density. A model based on boundedness, the density of roads at
the 2000 m scale, and the amount of prairie dog colonies at the 200 m scale explained 73% of the variance in prairie dog burrow
density. However, a non-linear model including boundedness and the squared term of road density at the 2000 m scale had the
lowest AIC value of all linear and non-linear models, indicating a possible threshold effect of urbanization on prairie dog density.
Urbanization may have several implications for prairie dog persistence. Increased prairie dog density in urbanized landscapes may
be related to the Refuge effect, i.e. decreased predator abundance. If higher prairie dog density increases competition for available
resources, habitat quality may decline leading to population decline in highly urbanized landscapes. Furthermore, if dispersal is
reduced in urbanized landscapes, then these colonies may not be recolonized after local extirpation from plague epizootics. Alternatively, urbanized colonies may be effectively isolated from plague vectors and reservoir hosts, which could result in a lower frequency of plague epizootics when compared to non-urbanized colonies.
# 2003 Elsevier Ltd. All rights reserved.
Keywords: Black-tailed prairie dogs; Density; Urbanization; Roads; Landscape context
1. Introduction
Habitat loss and fragmentation alter the spatial distribution of habitats and resources for native animals
and also affect the context in which native habitat remnants occur (Collinge, 1996; Mazerolle and Villard,
1999; Collinge et al., 2003). In rapidly urbanizing landscapes such as the western USA, once continuous habitats are increasingly surrounded by residential and
* Corresponding author. Tel.: +1-303-735-3242; fax: +1-303-4928699.
E-mail address: [email protected] (W.C. Johnson).
0006-3207/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0006-3207(03)00165-4
commercial development (e.g. Bolger et al., 2000). Such
changes in landscape context may have variable effects
on native species. For example, in Boulder, Colorado,
USA, small mammal abundance decreased as urbanization increased (Bock et al., 2002). In the same landscape, however, butterfly and grasshopper abundance
were unrelated to the amount of urbanization (Craig et
al., 1999; Collinge et al., 2003). Despite a growing
number of studies focused on changing landscape context and species persistence, responses of most native
organisms are not widely understood.
Prairie dogs (Cynomys spp.) are considered keystone
species (sensu Paine, 1969) and ecosystem engineers
488
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
(sensu Jones et al., 1994; Ceballos et al., 1999) of shortand mixed-grass prairie ecosystems of western North
America because of their disproportionately large effect
on grassland ecosystem structure and function (Whicker
and Detling, 1988; Kotliar et al., 1999, 2000). Many
different species including badgers (Taxidea taxus),
coyotes (Canis latrans), swift foxes (Vulpes velox),
prairie rattlesnakes (Crotalus viridis), ferruginous hawks
(Buteo regalis), golden eagles (Aquila chrysaetos), and
black-footed ferrets (Mustela nigripes), prey upon
prairie dogs, and prairie dog burrows provide shelter for
many invertebrates, reptiles, amphibians, burrowing
owls (Athene cunicularia), and small mammals (Koford,
1958; Agnew et al., 1986; Desmond and Savidge, 1996;
Goodrich and Buskirk, 1998; Kotliar et al., 1999; Kretzer and Cully, 2001).
Black-tailed prairie dogs (Cynomys ludovicianus) historically inhabited approximately 3 million ha in Colorado (Gillette, 1919; as cited in Colorado Department
of Natural Resources, 2000) and now only occupy
approximately 120,000 ha (Colorado Department of
Natural Resources, 2000). Range-wide, disease, competition with cattle ranching operations, agriculture, and
most recently urban and suburban development have
reduced prairie dogs to less than 2% of their original
abundance (Miller et al., 1990, 1994; American Society
of Mammalogists, 1998), and prairie dogs now occupy
less than 1% of the area in their estimated historical
geographic range (Gober, 2000). Poisoning and shooting of prairie dogs by ranchers, and agricultural conversion of habitat are responsible for the majority of the
prairie dog’s decline (Miller et al., 1990, 1994). Furthermore, the introduced sylvatic plague (Yersinia pestis), kills up to 99% of prairie dogs in infected colonies
(Cully et al., 1997; Cully and Williams, 2001; Biggins
and Kosoy, 2001). The earlier factors, in combination
with habitat loss from urbanization, helped the blacktailed prairie dog reach ‘‘warranted, but precluded’’
status by the United States Fish and Wildlife Service
(Gober, 2000).
Prairie dog colonies in Boulder County are located on
the Colorado Piedmont, which is located at the transition between the Great Plains and the Front Range of
the Rocky Mountains. Along with the rest of the Colorado Front Range corridor, human population in
Boulder County has rapidly increased, which exerts
pressure to develop native grasslands. Urbanization and
agriculture on the Colorado Piedmont have created
areas dominated by human activity within once continuous grassland, thereby ‘‘perforating’’ the landscape
(Forman, 1995; Collinge and Forman, 1998). The City
of Boulder Open Space and Mountain Parks Department (OSMP) has created a ‘‘greenbelt’’ of preserved
grassland properties (12,000 ha in 440 parcels) around
the city of Boulder to preserve native biodiversity along
with traditional land uses such as cattle ranching and
agriculture. However, these properties already exist
within a highly perforated prairie ecosystem (Berry et
al., 1998; Collinge and Forman, 1998; Bock et al., 2002).
As a result, the majority of prairie dog colonies in
Boulder County are on city- and county-owned land.
Many of the remaining prairie dog colonies in Boulder
have been displaced from their previous habitat and are
now located on roadsides and on leftover portions of
new urban developments. These ‘‘urban’’ prairie dog
colonies are uncharacteristic of prairie dog colonies
found in continuous grassland. However, it is unknown
whether these urban prairie dog colonies can survive
over the long term.
We indirectly measured the effects of landscape perforation on prairie dog colonies by studying prairie dog
colony characteristics in the short- and mixed-grass
prairie ecosystems of Boulder County. Our research
goal was to uncover relationships between two dependent variables: the density of active burrow entrances
(burrow density) and prairie dog density, and multiple
independent variables: colony area and the amount of
urbanization, roads, and other prairie dog colonies in
the surrounding landscape at three spatial scales. We
used correlation analyses to determine which independent variables were related to prairie dog density and
Akaike’s Information Criterion (AIC—Burnham and
Anderson, 1998) to determine the most parsimonious
models predicting burrow density using the above independent variables. We predicted that the amount of
urbanization, roads, and prairie dog colonies in the
surrounding landscape would significantly affect prairie
dog density. We also tested post-hoc for non-linear
relationships between burrow density and two indicators of urbanization—percent urbanization and road
density at all three spatial scales.
2. Methods
2.1. Study area
The city of Boulder is located in north central Colorado between the Great Plains to the east and the Front
Range of the Rocky Mountains to the west. All study
sites were located between 40.175 and 39.845 N Latitude and 105.320 and 105.079 W Longitude at an
average elevation of 1645 m. Approximately 1,200 ha of
active prairie dog colonies lie within the study area
(Fig. 1). Colony size ranges from 0.5 to 100 ha with an
average colony size of 8 ha (Johnson, unpublished
data). Prairie dog colonies in Boulder are located in
short- and mixed-grass prairies, all of which were historically grazed by bison, and more recently cattle.
Short-grass sites are dominated by Agropyron smithii
(western wheatgrass), Bouteloua gracilis (blue grama),
Buchloe dactyloides (buffalo grass), Artemisia frigida
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
489
Fig. 1. Prairie dog colonies within the perforated landscape of Boulder, Colorado.
(pasture sagebrush), and Plantago patagonica (woolly
plantain), and mixed-grass sites are dominated by Bouteloua gracilis (blue grama), Bouteloua curtipendula
(side-oats grama), Liatris punctata (blazing star), Artemisia ludoviciana (prairie sage), and Aster falcatus (aster).
2.2. Prairie dog density and burrow density
We estimated the density of prairie dogs using two
methods: the belt-transect method to calculate the density
of active burrow entrances (hereafter ‘‘burrow density’’)
(Biggins et al., 1993; Powell et al., 1994; Van Druff et
al., 1996), and the visual count method to calculate
prairie dog density (Fagerstone and Biggins, 1986;
Menkens et al., 1990; Menkens and Anderson, 1993;
Powell et al., 1994). We strategically chose a sample of
prairie dog colonies that spanned the full range of
landscape contexts. We sampled burrow density in the
months of June–August 2000, and September
2001–January 2002 using 350 m belt transects (Biggins
490
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
et al., 1993; Powell et al., 1994; Van Druff et al., 1996).
We randomly placed belt transects within each colony.
We included all three burrow opening types as classified
by Hoogland (1995) in estimates of burrow density:
burrow entrances with no mound, dome craters, and
rim craters. Active burrow entrances are characterized
as having new scat (Biggins et al., 1993; Hoogland,
1995). We calculated burrow density, area, and
‘‘boundedness’’ (see later definition) for 36 colonies in
2000 and 40 colonies in 2001. In 2000, we randomly
placed 2, 350 m belt-transects within 36 colonies and
counted the number of active burrow entrances. In 2001
we randomly placed 10, 3x50 m belt-transects within 40
colonies and recorded all active burrow entrances. We
used the number of active burrow entrances per hectare
to estimate burrow density in these colonies. We analyzed the 2000 and 2001 datasets separately because
boundedness, burrow density, and area change from
year to year (Hoogland, 1995).
We quantified prairie dog density in 15 colonies using
timed visual counts (Fagerstone and Biggins, 1986;
Menkens et al., 1990; Powell et al., 1994) during the
months of September 2001–January 2002. We strategically selected 15 out of the 40 colonies to represent the
range of boundedness values. Our purpose was to
determine the relationship between prairie dog density
and burrow density (Biggins et al., 1993; Powell et al.,
1994; Reading and Matchett, 1997). We counted prairie
dogs using 1024 mm binoculars at a distance of 100 m
from the colony perimeter. Depending on access to each
colony, we remained in a vehicle or on a camping chair
for the duration of the count. Because we were unable
to count every prairie dog in each colony, we staked out
between 1 and 8, 5050 m grids within each of 15
colonies prior to surveying. We used one grid for very
small colonies (colonies that could only fit one grid) and
up to eight for the largest colonies. We counted the
number of prairie dogs in each grid three times an hour,
three hours a day, for three days in a row. Counts at
each colony were performed between 09:00 and 12:00 h
on days with no precipitation and temperatures above
10 C. We arrived at least 15 min before 09:00 to allow
the prairie dogs to acclimate to human presence. To
estimate prairie dog density of each colony we calculated the maximum average count of prairie dogs. For a
given colony, the maximum average is calculated by
adding the maximum number of prairie dogs counted in
each grid over the three day sampling period and dividing that sum by the number of grids at that colony.
Prairie dog density is calculated by dividing the maximum average by the area of one grid.
With the exception of the capture/mark/recapture
technique, calculating the maximum average of visual
count data is considered the most accurate estimate
of actual prairie dog density (Severson and Plumb,
1998).
2.3. Landscape context
In 2000 and 2001, we determined the boundedness of
36 and 40 colonies, respectively, on a scale of 0–5
depending on the colony’s surroundings. Boundedness
describes the immediate surroundings of a colony and
may be considered an estimate of the relative permeability of the colony to emigration and immigration. We
used the amount of unsuitable habitat surrounding each
colony to calculate Boundedness. Unsuitable habitat
includes most non-grassland habitats: the foothills of
the Front Range, lakes and reservoirs, dirt and paved
roads, freeways, and urban and industrial developments. In effect, this is an estimation of the amount of
urban and industrial development and roads directly
adjacent to each prairie dog colony. For example, if the
colony was surrounded on all sides by short- or mixedgrass prairie, then the boundedness value was zero
(Fig. 2). Conversely, if the colony was surrounded on all
sides by high-density urban development, then the
boundedness value was five. Boundedness was measured
for each colony by the same individual to ensure consistency between observations.
We further determined landscape context for 22
prairie dog colonies using remotely sensed data and GIS
analyses. In November 2000 and 2001, we obtained a
GIS layer of all colonies found on Boulder City and
County Open Space properties. We used Global Positioning Systems (GPS, Trimble GeoExplorer II) to map
the perimeter of prairie dog colonies that were not
found on City or County of Boulder Open Space properties. We merged these data into one complete GIS
layer incorporating all prairie dog colonies found in
Boulder County over both the 2000 and 2001 seasons.
In 2001, there were approximately 260 active colonies in
Boulder County.
We quantified three landscape features in relation to
prairie dog colonies at three spatial scales: urbanization,
road density, and the amount of other prairie dog colonies at 200, 1000, and 2000 m from the perimeter of
each colony. First, we digitized urban, suburban, and
industrial development (hereafter, urbanization) in
Boulder County using 62 digital orthoquad images
(DOQs) with 1 m resolution. We defined urbanization
as any anthropogenic landscape feature, including residential and industrial buildings, parking lots, urban
vegetation, and roads. Second, we calculated road density using a GIS layer for all roads that occurred within
Boulder County. Third, we calculated the area of other
prairie dog colonies using the prairie dog GIS layer. We
considered the area of other prairie dog colonies within
the surrounding landscape as an indication of colony
isolation, i.e. more isolated colonies had lower areas of
prairie dog colonies in the surrounding landscape.
We strategically chose 22 out of the 40 colonies on which
to perform landscape context analyses. We stratified our
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
491
Fig. 2. Examples of boundedness values (0–5) for six colonies. Colonies are circumscribed by the thin black lines in each figure. We classified a
colony as ‘‘1’’ if there was unsuitable habitat on one side of the colony, ‘‘2’’ if there was unsuitable habitat on two sides of the colony, and so on.
Colonies with a boundedness value of ‘‘5’’ were completely surrounded by a high density of buildings and roads on all sides. Aerial photos taken in
1999.
sample to include a similar number of colonies in each
of three boundedness zones: low=0–1, medium=2–3,
high=4–5. We did not randomly select colonies because
highly urbanized colonies were fewer in number than
other colonies. Using ArcView GIS (ArcView v3.2,
ESRI Inc., Redlands, CA), we created a buffer around
each of the 22 colonies at three spatial scales (200, 1000,
and 2000 m). This created three zones (polygons)
extending outward 200, 1000, and 2000 m from the
perimeter. The largest spatial scale of 2 km was chosen
because intercolony prairie dog dispersal rarely exceeds
2–3 km (Garrett and Franklin, 1988; Hoogland, 1995).
We then used the ‘‘Clip’’ function in the GeoProcessing
Wizard extension (ArcView v.3.2, ESRI Inc., Redlands,
CA, USA) to select all urbanization, roads, and other
prairie dog colonies that fell within the boundary of
each buffer zone. We then created a new data layer
containing the clipped features and their associated
lengths or areas. We calculated percent urbanization,
road density, and percentage of other prairie dog colonies within each buffer by dividing the area (urbanization and prairie dog colonies) or length (roads) of the
clipped theme by the area of the buffer theme and multiplying by 100. In this way, we could determine at
which of the three spatial scales each of these three
independent variables affected the dependent variables:
prairie dog density and burrow density.
2.4. Statistical analysis
For the 2000 field season, we examined the association
between boundedness and burrow density using Pearson
correlation (n=36) (Zar, 1999). For the 2001 field season, we first analyzed relationships between boundedness, prairie dog colony area, prairie dog density (n=15)
and burrow density (n=40) using Pearson and Spearman rank correlation. Second, we compared all three
scales of each landscape context variable to burrow
density (n=22) using Spearman rank correlation. Third,
we used AIC to determine the most parsimonious models predicting burrow density (Burnham and Anderson,
1998; Roach et al., 2001). We combined 11 independent
variables to create a number of regression models. The
predictor variables we used in the AIC analysis were: (1)
area of colony, (2) 200 m % urbanization, (3) 1000 m %
urbanization, (4) 2000 m % urbanization, (5) 200 m %
road density, (6) 1000 m % road density, (7) 2000 m %
road density, (8) 200 m % prairie dog colonies, (9) 1000
m % prairie dog colonies, (10) 2000 m % prairie dog
colonies, and (11) boundedness. We did not test all
combinations of all 11 independent variables using AIC
because it is best to keep the number of candidate
models to a minimum by using previous knowledge of
the system (Burnham and Anderson, 1998). For example, we excluded models that used more than one scale
of urbanization, roads, or prairie dog colonies together
because each landscape variable was not independent
across scales (see Section 3). Unlike hypothesis testing,
AIC does not use an arbitrary significance level (i.e.
a=0.05) to test for model significance. Instead, using a
log-likelihood approach, AIC tests a set of a priori
selected models to find which combinations of the selected independent variables most parsimoniously predict
the dependent variable (Burnham and Anderson, 1998).
AIC also discounts models that use a large number of
variables because a model’s predictive value decreases
with too many predictor variables. The goal of AIC is
not to determine the significance of any particular model,
but to determine which of the tested models is best used
for inference and further exploration. Because Bock et
492
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
al. (2002) found a threshold effect of urbanization on
small mammal abundance we tested post-hoc for non-linear effects of urbanization and road density on prairie dog
density. We used the three most parsimonious linear
models (see Table 2) to substitute a squared term of urbanization or road density at each of the three spatial scales
to test for non-linear effects of these variables. We also
squared and tested each individual scale of urbanization
and road density for non-linear effects on burrow density.
We calculated the standard error for each regression
and used this standard error to calculate AIC for each
regression (Burnham and Anderson, 1998). Each regression has its own AIC value and the lowest AIC value is
the model that most parsimoniously explains the variance in the data (Burnham and Anderson, 1998).
Because of low sample size (n=22), we calculated the
corrected AIC (AICc) for all models (Burnham and
Anderson, 1998). To compare models, we computed
AICi, the difference in AIC values between model i and
the most parsimonious model. The best model has the
lowest AICc value and therefore a AIC value of zero.
Models with AIC < 4 should be considered useful
candidates for explaining variance in the dependent
Table 1
Landscape context variables and partial correlation coefficients with
active burrow entrance density
Independent variable
n
Prairie dog density (prairie dogs/ha)
Area of colony (ha)
Boundedness
% Urbanization (200 m scale)
% Urbanization (1000 m scale)
% Urbanization (2000 m scale)
Road density (200 m scale)
Road density (1000 m scale)
Road density (2000 m scale)
Prairie dog colonies (200 m scale)
Prairie dog colonies (1000 m scale)
Prairie dog colonies (2000 m scale)
15
40
40
22
22
22
22
22
22
22
22
22
*P< 0.05. **P <0.01. ***P <0.001.
r
0.60*
0.37*
0.81***
0.63**
0.67***
0.65***
0.43*
0.63**
0.67***
0.14
0.07
0.003
variable (Burnham and Anderson, 1998). Models with a
AIC42 are considered the most parsimonious.
Lastly, we ranked each model based on its calculated
‘‘Akaike Weight’’ value, a measure of relative likelihood
(Burnham and Anderson, 1998).
All spatial analyses were conducted with ArcView
v.3.2 (ESRI, Redlands, CA, USA), and all statistical
analyses were performed with SAS (Statistical Analysis
Software v.8.2, Cary, NC, USA).
3. Results
A full Pearson correlation matrix analysis of all independent variables revealed that all scales of urbanization were significantly correlated with each other (P
< 0.0001), all scales of road density were significantly
correlated with each other (P < 0.005), and all scales of
the percentage of the landscape covered by prairie dog
colonies were significantly correlated with each other.
Furthermore, all scales of road density were correlated
with boundedness (P < 0.05), and within the 2000 m
scale road density was negatively correlated with the
percentage of the landscape covered with prairie dog
colonies (P < 0.05) and colony area (P < 0.05).
Burrow density for the 22 colonies ranged from 100 to
674 burrows/ha, and prairie dog density ranged from 32
to 120 prairie dogs/ha. Burrow density and prairie dog
density were positively correlated (Fig. 3; n=15; Pearson correlation, r=0.80, P < 0.001; Spearman rank
correlation, r=0.60, P < 0.05). In 2000, burrow density
was positively correlated with boundedness (n=36;
Pearson correlation, r=0.34, P < 0.05). In 2001, burrow
density and prairie dog density were both positively
correlated with boundedness (Fig. 4a and b, Spearman
correlation, n=40 & 15, r=0.66 and 0.81, respectively,
P < 0.001). Colony area was negatively correlated with
boundedness (Fig. 4c; Spearman correlation, n=40;
r= 0.37; P=0.02), but was not correlated with burrow
density (P=0.11).
Table 2
AIC analysis: the eight linear regression models that best explained prairie dog burrow density across a gradient of urbanization (See text for
explanation of Akaike’s information criterion)
Model No.
Variables
R2
adj R2
AICc
AICc
Akaike
weight
1
2
3
4
5
Boundedness, road density at 2000 m scale, prairie dog colonies at 200 m scale
Boundedness, road density at 2000 m scale
Road density at 2000 m scale, prairie dog colonies at 200 m scale
Boundedness
Area of colony, Boundedness, road density at 2000 m scale, prairie dog colonies
at 200 m scale
Area of colony, Boundedness, road density at 2000 m scale
Boundedness, road density at 2000 m scale, prairie dog colonies at 2000 m scale
Area of colony, road density at 2000 m scale, prairie dog colonies at 200 m scale
0.7251
0.6762
0.6655
0.5753
0.7314
0.6793
0.6421
0.6303
0.554
0.6682
212.9
213.1
213.8
216.1
216.2
0*
0.2086*
0.9212*
3.159
3.343
0.1913
0.1724
0.1207
0.0394
0.0397
0.6766
0.6763
0.6725
0.6227
0.6224
0.6179
216.5
216.5
216.8
3.576
3.596
3.857
0.0320
0.0317
0.0278
6
7
8
*=Models with AICc <2 were considered the best approximating models predicting prairie dog burrow density.
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
493
Fig. 3. Correlation between prairie dog density and active burrow entrance density (r=0.60, P <0.05). Data for 2001, n=15.
Burrow density was positively correlated with urbanization at the 200, 1000, and 2000 m spatial scales
(Table 1; Fig. 5a; n=22, r=0.63, 0.67, 0.65; P < 0.005,
0.001, 0.001, respectively). Similarly, burrow density was
positively correlated with road density at all three spatial
scales (Table 1; Fig. 5b; n=22, r=0.43, 0.63, 0.67; P
< 0.05, 0.005, 0.001, respectively). However, the percentage of the landscape covered by prairie dog colonies was
not correlated with burrow density at any scale (Table 1;
Fig. 5c; r=0.14, 0.07, 0.003, respectively; P > 0.5).
In the AIC analysis, the most parsimonious linear model
was based on boundedness, the density of roads at the
2000 m scale, and the percentage of the landscape covered
by prairie dog colonies at the 200 m scale. This model
explained 73% of the variance in burrow density (Table 2).
The second best linear model included boundedness and
road density at the 2000 m scale, and explained 68% of the
variance in burrow density (Table 2). The third most parsimonious linear model was based on road density at the
2000 m scale and the percentage of the landscape covered
by prairie dog colonies at the 200 m scale, and explained
67% of the variance in burrow density (Table 2).
The post-hoc non-linear AIC analysis resulted in
three models with AIC less than 2—all of which had
lower AICc values than the most parsimonious linear
model. The first model included boundedness and road
density at the 2000 m scale squared (R2=0.73;
AICc=209.9). The second model included boundedness,
road density at the 2000 m scale squared, and the percentage of the landscape covered in prairie dog colonies
at the 200 m scale (R2=0.75; AICc=210.4). The third
model included just the density of roads at the 2000 m
scale squared (R2=0.65; AICc=211.8).
4. Discussion
Urbanization and roads in the immediate surroundings of prairie dog colonies (boundedness) and in the
surroundings of prairie dog colonies at larger spatial
scales were positively correlated with, and best predicted
densities of black-tailed prairie dogs. Conversely, lower
prairie dog density within colonies appeared to be associated with the presence of nearby prairie dog colonies.
Further, our results appear to show a positive non-linear trend of road density on prairie dog density (Fig. 5).
Interestingly, non-linear models with a squared term of
percent urbanization did not perform better than the
most parsimonious linear model. However, models with
a squared term of road density did perform better than
the most parsimonious linear model. In contrast to what
Bock et al. (2002) found with rodent abundance, there
appears to be a positive non-linear relationship in the
effect of urbanization—as measured with road density—
on the density of prairie dogs in Boulder County.
4.1. Burrow density and prairie dog density
Colonies with more active burrow entrances had more
prairie dogs. The average burrow entrance density at
our study sites was 255 active burrow entrances/ha, and
the average number of prairie dogs was 68 prairie dogs/
ha. The averages and ranges in our study area were
higher than the ranges of 10–250 burrows/ha and 10–35
prairie dogs/ha that have been documented in other
studies (Koford, 1958; Reading et al., 1989; Powell et
al., 1994). There is an equivocal relationship between
active burrow density and prairie dog density (Biggins et
494
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
al., 1993; Powell et al., 1994; Reading and Matchett,
1997; VanHorne et al., 1997). Unlike the majority of
studies using similar methodology, we found a significant correlation between these variables. Previous
studies may have used different protocols to sample
active burrow entrance densities. Prairie dog burrows
usually have more than one entrance and there is more
than one type of burrow entrance (Hoogland, 1995).
Therefore, specific methods by which burrows are
counted may contribute to the difference among studies
in the relationship between active burrow density and
prairie dog density (Reading et al., 1989; Van Horne et
al., 1997; Severson and Plumb, 1998). Furthermore,
because other studies did not find such a large range of
burrow density values and were located in areas without
urbanization, it is possible that the significant relationship between burrow density and prairie dog density
depends on landscape context.
Fig. 4. (a) Positive correlation between burrow density (burrows/ha)
and boundedness (r=0.66, P <0.001). (b) Positive correlation
between prairie dog density (prairie dogs/ha) and boundedness
(r=0.81, P <0.001). (c) Negative correlation between colony area and
boundedness (r= 0.37, P<0.05).
Fig. 5. Spearman correlations of the landscape context of prairie dog
colonies at the 2000 m scale. Burrow density increased with (a) percent
urbanization (P < 0.001), (b) percent road density (P <0.001), and (c)
was not correlated with percent of the landscape covered in other
prairie dog colonies (P > 0.5).
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
4.2. Landscape context of Boulder County
In Boulder County, mixed-grass prairies, along with
the prairie dog colonies within these grasslands, have
been perforated by urbanization and agriculture forcing
prairie dogs to survive wherever possible. There are
numerous small, bounded colonies along roadsides and
both large and small colonies in areas of continuous
grassland. Historically, soil structure, topography, and
vegetation structure restricted colony expansion (Koford,
1958; Reading and Matchett, 1997). Currently, urbanization and other forms of human development restrict
prairie dog colony size and spatial distribution more than
any other factor, especially in Boulder County.
In Boulder County, the effects of increased urbanization on other grassland species are variable. Increased
urbanization was negatively correlated with the abundance of grassland nesting songbirds (Haire et al.,
2000), small mammals (Bock et al., 2002), and wintering
raptors (Berry et al., 1998), but not correlated with the
abundance of butterflies (Collinge et al., 2003), and
grasshoppers (Orthoptera) (Craig et al., 1999). Deer
mice (Peromyscus maniculatus), prairie voles (Microtus
ochrogaster), hispid pocket mice (Chaetodipus hispidus),
wintering raptors, and grassland nesting songbirds were
found to decrease abruptly in abundance at very small
amounts of urbanization in the surrounding landscape
(Berry et al., 1998; Haire et al., 2000; Bock et al., 2002).
The authors of these studies hypothesized that a critical
landscape threshold (Andren, 1994; With and Crist,
1995) exists at 5–7% urbanization of the landscape
where the abundance of these species sharply decreases.
Interestingly, the opposite effect occurred in our study
on prairie dogs: prairie dog density increased non-linearly with increasing boundedness and roads. The
increase in prairie dog density that we observed may be
due to a ‘‘refuge effect’’ in urbanized landscapes. For
example, predation on nests of songbirds was significantly higher at greater distances from recreational
trails in Boulder County riparian areas, presumably due
to human presence on recreational trails (Miller and
Hobbs, 2000). Similarly, many potential prairie dog
predators may decline in urbanized landscapes enabling
prairie dog colonies to achieve higher densities.
Few other studies have looked at landscape context
effects on prairie dog density. Reading and Matchett
(1997) found that in the relatively non-urbanized grasslands of Montana distance to roads did not affect
prairie dog density or the area of colonies. In our study,
road density strongly affected prairie dog density; however, indices of landscape context were calculated differently in these two studies, and average colony size in
Montana is much greater than in Boulder County.
In a study of the Mexican prairie dog (Cynomys
mexicanus), Trevino-Villareal and Grant (1998) found
lower burrow density on small colonies that had been
495
fragmented by agriculture. The authors speculated that
habitat loss and fragmentation were responsible for the
local extirpation of 6 out of 88 colonies because of
decreased connectivity of the landscape. If smaller, more
fragmented colonies are in fact more vulnerable to
extinction, then the smaller, higher-density, and more
perforated colonies in our study may have a higher
extinction risk than colonies in less perforated landscapes.
4.3. Habitat quality and demography of prairie dogs
Increased prairie dog density could result in multiple
demographic changes. Higher density of animals does
not necessarily indicate higher habitat quality (Van
Horne, 1983), especially for prairie dogs in Boulder
County. In a non-urbanized landscape, Gunnison’s
prairie dog (Cynomys gunnisoni) was observed to have
similar densities at good and poor habitat quality sites,
but prairie dogs at the poor habitat quality site had
lower body mass, delayed sexual maturity, and delayed
dispersal when compared to a site with higher habitat
quality (Rayor, 1985). Similarly, at an urban prairie dog
colony in Boulder County, adult males, and adult and
juvenile females had significantly lower body mass than
the same age groups at a ‘‘rural’’ colony (Dawson,
1991). Dawson’s (1991) urban site was surrounded by
buildings and the rural colony was located in relatively
undisturbed grassland. Interestingly, Dawson (1991)
found no difference in prairie dog densities between the
urban and rural prairie dog colonies, contrary to our
study. However, he sampled only one urban and one
rural colony. Even without higher densities, if urban
prairie dogs have decreased body mass (Dawson, 1991)
and prairie dogs in poorer habitat quality sites have
decreased body mass (Rayor, 1985), then prairie dogs in
urban colonies may have decreased body mass due to
poor habitat quality. More research is necessary to
determine if high-density prairie dog colonies in urbanized landscapes actually have lower food resources
than colonies in non-urbanized landscapes.
In highly urbanized colonies, prairie dog density may
also be related to dispersal rates. In colonies not bounded
by urbanization, dispersal rates increased as available
food resources decreased (Garret and Franklin, 1988).
Voles experience what has been coined the ‘‘fence effect’’
(Krebs et al., 1969). When voles are fenced (bounded)
into a habitat patch vole density increases to an abnormal
high followed by a severe population crash (Krebs et al.,
1969). If prairie dog dispersal is reduced in urban colonies
because of barriers such as roads and buildings, and
increased density of prairie dogs decreases food resources
in urban colonies, then prairie dogs may have lower dispersal rates in areas of low food resources. Over the long
term, high-density colonies may not be able to sustain
high amounts of herbivory, and habitat quality may
become degraded leading to population decline.
496
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
4.4. Plague and prairie dogs
In addition to habitat loss and perforation effects on
prairie dog colonies, sylvatic plague may now have created
metapopulations of prairie dogs (Roach et al., 2001).
Prairie dog colonies are not thought to have experienced
periodic extinctions before the introduction of plague
(Cully and Williams, 2001; Roach et al., 2001). Colonies
are now periodically extirpated by plague and later
recolonized by dispersers from nearby colonies (Cully and
Williams, 2001; Roach et al., 2001). Colonies with higher
prairie dog density appear to have higher plague transmittance rates than less dense colonies (Cully and Williams,
2001). If plague were to appear at an urbanized colony
complete extirpation would likely ensue. However, plague
vectors may not be able to carry plague to urbanized
colonies. For example, during the 1994 plague epizootic
in Boulder County, many of the more urbanized colonies were spared from plague (Gershman, 1996).
Colony size may also affect transmission rates of plague
between colonies. Cully and Williams (2001) found that
large colonies less than 3 km apart were more likely to
experience plague during an outbreak than small colonies.
They suggested that small colonies greater than 3 km
from other colonies tended to persist longer when plague is present in the landscape. In Boulder County, few
colonies are disconnected by more than 3 km; colonies
are dispersed throughout the landscape like steppingstones. Therefore, the limiting factor affecting disease
transmittance in Boulder County may not be whether the
colonies are isolated by distance, but whether they are isolated by their surroundings (i.e. their landscape context).
In highly urbanized landscapes, there may be
decreased dispersal success and decreased intraspecific
plague transmission. Cully and Williams (2001) suggested that plague may be carried to colonies interspecifically by animals such as coyotes, deer mice, and
raptors. Interspecific disease transmission may also be
decreased by urbanization; thus, transmitting plague to
these urban prairie dog colonies could prove difficult.
4.5. Conservation Implications
Prairie dog colonies in urbanized areas of Boulder
County have densities of prairie dogs above average
compared to colonies in less developed areas. The longterm implications of such high prairie dog density are
unknown, although one can imagine effects on habitat
quality, resource abundance, inbreeding rate, dispersal
rate, and plague occurrence and transmission. With no
room for colony expansion, high prairie dog density in
urbanized areas could create increased competition for
available food and space, lower dispersal rates, and possibly increase inbreeding depression. Movement corridors could be constructed to provide interaction between
rural and urban prairie dog colonies; however, corridors
could also increase the movement of plague reservoirs
and vectors, thereby increasing the risk of plague to
urban colonies (Hess, 1996). Furthermore, in the event of
plague, high density colonies may experience increased
plague transmission rates and decreased recolonization
rates jeopardizing the survival of urban colonies. However, the effects of urbanization on plague transmission
will depend on what species, in addition to prairie dogs,
may transmit plague between colonies, and how potential dispersal barriers, such as roads, affect these species.
Acknowledgements
We thank the City of Boulder Open Space and
Mountain Parks Department for allowing us to conduct
this research on their properties and supplying us with
GIS layers of Boulder; our field assistants L. McCauley,
A. Benson, and H. Valdez; and C. Ray for her limitless
statistical advice and editorial comments. We appreciate
helpful comments from D. Armstrong and C. Bock on
earlier drafts of this manuscript. This research was completed in partial fulfillment of a MS degree to W. Johnson. Financial support was provided in part by grants
from EPA (R-82909101) and NSF (DEB-0224328).
References
Andren, H., 1994. Effects of habitat fragmentation on birds and
mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71, 355–366.
Agnew, W., Uresk, D.W., Hansen, R.M., 1986. Flora and fauna
associated with prairie dog colonies and adjacent ungrazed mixedgrass prairie in western South Dakota. Journal of Range Management 39, 135–139.
American Society of Mammalogists, 1998. American Society of Mammalogists’ resolution on the decline of prairie dogs and the grassland
ecosystem in North America. Journal of Mammalogy 79, 1447–1448.
Berry, M.E., Bock, C.E., Haire, S.L., 1998. Abundance of diurnal
raptors on open space grasslands in an urbanized landscape. The
Condor 100, 601–608.
Biggins, D., Miller, B., Hanebury, L., Oakleaf, B., Farmer, A., Crete, R.,
Dood, A., 1993. A system for evaluating black-footed ferret habitat.
In: Management of Prairie Dog Complexes for Black-footed Ferret
Reintroduction. Oldemeyer, J., Biggins, D., Miller, B., Crete, R.
(Eds.), US Fish and Wildlife Service Biological Report 13. pp. 73–92.
Biggins, D.E., Kosoy, M.Y., 2001. Influences of introduced plague on
North American mammals: implications from ecology of plague in
Asia. Journal of Mammalogy 82, 906–916.
Bock, C.E., Vierling, K.T., Haire, S.L., Boone, J.D., Merkle, W.W.,
2002. Patterns of rodent abundance on open-space grasslands in
relation to suburban edges. Conservation Biology 16, 1653–1658.
Bolger, D.T., Suarez, A.V., Crooks, K.R., Morrison, S.A., Case, T.J.,
2000. Arthropods in urban habitat fragments in southern California:
area, age, and edge effects. Ecological Applications 10, 1230–1248.
Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference. A Practical Information-theoretic Approach. Springer-Verlag
New York Inc, New York, NY.
Ceballos, G., Pacheco, J., List, R., 1999. Influence of prairie dogs
(Cynomys ludovicianus) on habitat heterogeneity and mammalian
diversity in Mexico. Journal of Arid Environments 41, 161–172.
W.C. Johnson, S.K. Collinge / Biological Conservation 115 (2004) 487–497
Collinge, S.K., 1996. Ecological consequences of habitat fragmentation: implications for landscape architecture and planning. Landscape and Urban Planning 36, 59–77.
Collinge, S.K., Forman, R.T.T., 1998. A conceptual model of land
conversion processes: predictions and evidence from a microlandscape experiment with grassland insects. Oikos 82, 66–84.
Collinge, S.K., Prudic, K.L., Oliver, J.C., 2003. Effects of local habitat
characteristics and landscape context on grassland butterfly diversity. Conservation Biology 17, 178–187.
Colorado Department of Natural Resources. 2000. Black-tailed
Prairie Dog Study of Eastern Colorado. Prepared by EDAW, Inc.
Fort Collins, CO.
Craig, D.P., Bock, C.E., Bennett, B.C., Bock, J.H., 1999. Habitat
relationships among grasshoppers (Orthoptera:Acrididae) at the
western limit of the Great Plains in Colorado. American Midland
Naturalist 142, 314–327.
Cully Jr., J.F., Barnes, A.M., Quan, T.J., Maupin, G., 1997. Dynamics
of plague in a Gunnison’s prairie dog colony complex from New
Mexico. Journal of Wildlife Diseases 33, 706–719.
Cully Jr., J.F., Williams, E.S., 2001. Interspecific comparisons of sylvatic plague in prairie dogs. Journal of Mammalogy 82, 894–905.
Dawson, R.E., 1991. Comparison of Body Mass and Demography
between Urban and Rural Populations of Black-tailed Prairie Dogs
(Cynomys ludovicianus). MS Thesis, University of Colorado at
Boulder.
Desmond, M.J., Savidge, J.A., 1996. Factors influencing burrowing
owl (Speotyto cunicularia) nest densities and numbers in western
Nebraska. American Midland Naturalist 136, 143–148.
Fagerstone, K.A., Biggins, D.E., 1986. Comparison of capture-recapture and visual indices of prairie dog densities in black-footed ferret
habitat. Great Basin Naturalist Memoirs 8, 94–98.
Forman, R.T.T., 1995. Land Mosaics: the Ecology of Landscapes and
Regions. Cambridge University Press, New York, NY.
Garrett, M.G., Franklin, W.L., 1988. Behavioral ecology of dispersal
in the black-tailed prairie dog. Journal of Mammalogy 69, 236–250.
Gershman, M.D., 1996. City of Boulder Open Space and Mountain
Parks Grassland Management, Black-tailed Prairie Dog Habitat
Conservation Plan. Available from: http://www.ci.boulder.co.us/
openspace/planning/pdogplan/pdogmain.htm.
Gober, P., 2000. 12 month administrative finding, black-tailed prairie
dog. Federal Register 65, 5476–5488.
Goodrich, J.M., Buskirk, S.W., 1998. Spacing and ecology of North
American badgers (Taxidea taxus) in a prairie-dog (Cynomys leucurus) complex. Journal of Mammalogy 79, 171–179.
Haire, S.L., Bock, C.E., Cade, B.S., Bennett, B.C., 2000. The role of
landscape and habitat characteristics in limiting abundance of
grassland nesting songbirds in an urban open space. Landscape and
Urban Planning 48, 65–82.
Hess, G., 1996. Disease in metapopulation models: implications for
conservation. Ecology 77, 1617–1632.
Hoogland, J.L., 1995. The Black-tailed Prairie Dog: Social Life of a
Burrowing Mammal. The University of Chicago Press, Chicago, IL.
Jones, C.G., Lawton, J.H., Shachak, M., 1994. Organisms as ecosystem engineers. Oikos 69, 373–386.
Koford, C.B., 1958. Prairie dogs, whitefaces, and blue grama. Wildlife
Monographs 3, 1–78.
Kotliar, N.B., Baker, B.W., Whicker, A.D., Plumb, G., 1999. A critical review of assumptions about the prairie dog as a keystone
species. Environmental Management 24, 177–192.
Kotliar, N.B., 2000. Application of the new keystone species concept
to prairie dogs: how well does it work? Conservation Biology 14,
1715–1721.
Krebs, C.J., Keller, B.L., Tamarin, R.H., 1969. Microtus population
biology: demographic changes in fluctuating populations of M.
ochrogaster and M. pennsylvanicus in southern Indiana. Ecology 50,
587–607.
497
Kretzer, J.E., Cully Jr., J.F., 2001. Effects of black-tailed prairie dogs
on reptiles and amphibians in Kansas shortgrass prairie. The
Southwestern Naturalist 46, 171–177.
Mazerolle, M.J., Villard, M., 1999. Patch characteristics and landscape context as predictors of species presence and abundance: a
review. Ecoscience 6, 117–124.
Menkens, G.E., Jr., Anderson, S.H., 1993. Mark-recapture and visual
counts for estimating population size of white-tailed prairie dogs.
In: Oldemeyer, J., Biggins, D., Miller, B., Crete, R. (Eds.), Management of Prairie Dog Complexes for Black-footed Ferret Reintroduction. US Fish and Wildlife Service Biological Report. pp. 67–72.
Menkens Jr., G.E., Biggins, D.E., Anderson, S.H., 1990. Visual counts
as an index of white-tailed prairie dog density. Wildlife Society
Bulletin 18, 290–296.
Miller, B., Wemmer, C., Biggins, D., Reading, R., 1990. A proposal to
conserve black-footed ferrets and the prairie dog ecosystem. Environmental Management 14, 763–769.
Miller, B., Ceballos, G., Reading, R., 1994. The prairie dog and biotic
diversity. Conservation Biology 8, 667–681.
Miller, J.R., Hobbs, N.T., 2000. Recreational trails, human activity,
and nest predation in lowland riparian areas. Landscape and Urban
Planning 50, 227–236.
Paine, R.T., 1969. A note on trophic complexity and community stability. American Naturalist 103, 91–93.
Powell, K.L., Robel, R.J., Kemp, K.E., Nellis, M.D., 1994. Aboveground counts of black-tailed prairie dogs—temporal nature and
relationship to burrow entrance density. Journal of Wildlife Management 58, 351–355.
Rayor, L.S., 1985. Effects of habitat quality on growth, age of first
reproduction, and dispersal in Gunnison’s prairie dogs (Cynomys
gunnisoni). Canadian Journal of Zoology 63, 2835–2840.
Reading, R.P., Beissinger, S.R., Grensten, J.J., Clark, T.W., 1989.
Attributes of black-tailed prairie dog colonies in north central
Montana, with management recommendations for the conservation
of biodiversity. In: Clark, T.W., Hinkley, D., Rich, T. (Eds.), The
Prairie Dog Ecosystem: Managing for Biological Diversity. Montana BLM Wildlife Technical Bulletin No. 2. pp. 13–28.
Reading, R.P., Matchett, R., 1997. Attributes of black-tailed prairie
dog colonies in north central Montana. Journal of Wildlife Management 61, 664–673.
Roach, J.L., Stapp, P., Van Horne, B., Antolin, M.F., 2001. Genetic
structure of a metapopulation of black-tailed prairie dogs. Journal
of Mammalogy 82, 946–959.
Severson, K.E., Plumb, G.E., 1998. Comparison of methods to estimate population of black-tailed prairie dogs. Wildlife Society Bulletin 26, 859–866.
Trevino-Villarreal, J., Grant, W.E., 1998. Geographic range of the
endangered Mexican prairie dog (Cynomys mexicanus). Journal of
Mammalogy 79, 1273–1287.
Van Druff, L.W., Bolen, E.G., San Julian, G.J., 1996. Management of
urban wildlife. In: Bookhout, T.A. (Ed.), Research and management techniques for wildlife and habitats, fifth ed. The Wildlife
Society, Bethesda, MD, pp. 507–530.
Van Horne, B., 1983. Density as a misleading indicator of habitat
quality. Journal of Wildlife Management 47, 893–901.
Van Horne, B., Schooley, R.L., Knick, S.T., Olson, G.S., Burnham,
K.P., 1997. Use of burrow entrances to indicate densities of Townsend’s ground squirrels. Journal of Wildlife Management 61, 92–101.
Whicker, A.D., Detling, J.K., 1988. Ecological consequences of prairie
dog disturbances: prairie dogs alter grassland patch structure,
nutrient cycling, and feeding-site selection by other herbivores.
BioScience 38, 778–785.
With, K.A., Crist, T.O., 1995. Critical thresholds in species responses
to landscape structure. Ecology 76, 2446–2459.
Zar, J.H., 1999. Biostatistical Analysis, fourth ed. Prentice Hall, Inc,
New Jersey.