Urban Maternity-Roost Selection by Big Brown Bats in
DANIEL J. NEUBAUM,
Department of Fish, Wildlife, and Conservation Biology, Colorado State University, Fort Collins, CO 80523, USA
KENNETH R. WILSON, Department of Fish, Wildlife, and Conservation Biology, Colorado State University, Fort Collins, CO 80523, USA
THOMAS J. O’SHEA, Fort Collins Science Center, United States Geological Survey, 2150 Centre Avenue, Building C, Fort Collins, CO 80526, USA
ABSTRACT Despite prevalent use of anthropogenic structures by bats and the associated implications for public health, management, and
bat conservation, very little quantitative information exists about urban roost characteristics and their selection by bats. During the summers of
2001 to 2004 we conducted fieldwork in Fort Collins, Colorado, USA, situated on the northern end of Colorado’s Front Range, to address
questions of roost selection by the big brown bat (Eptesicus fuscus). The city has experienced its greatest growth in the past half century, with its
population increasing by 30%in the last decade. Similar growth in new buildings has occurred, with the number of new housing permits issued
annually doubling in the past decade. We located 142 roosts using radiotelemetry or by citizen calls in response to a newspaper article and flyers.
To determine characteristics of roost selectivity by bats, we compared variables for known maternity roosts and randomly selected buildings at
microhabitat and landscape scales using logistic regression; we used an information theoretic approach to determine which variables were most
important. We considered 44 and 100 buildings in the microhabitat and landscape scale analyses, respectively. At the microhabitat scale
maternity roosts had exit points with larger areas that were higher from the ground and had warmer average temperatures than randomly
selected buildings. At the landscape scale distances to similarly categorized roosts were smaller, and urbanization variables such as lower
building density, higher street density, and lower traffic count density were most important. Results for variables important to urban-roosting
big brown bats were often analogous to studies that characterized maternity roosts found in tree snags and rock crevices. In addition, changes in
the landscape, not only in the form of anthropogenic structures but also in water availability and vegetation structure such as riparian forests,
may have led to population increases and range expansions of the big brown bat. Because big brown bats appear to selectively choose specific
combinations of characteristics found at maternity roosts, not all available structures can be considered suitable and exclusion from established
maternity roosts may negatively impact bat populations. ( JOURNAL OF WILDLIFE MANAGEMENT 71(3):728–736; 2007)
KEY WORDS anthropogenic structures, big brown bat, Eptesicus fuscus, information theory, landscape scale, maternity-roost
selection, microhabitat scale, urbanization.
All temperate zone species of bats need summer roosts with
appropriate microclimatic conditions for raising young
within commuting distance of foraging and watering sites
(Pierson 1998). Roosting requirements are directed by a
variety of factors (e.g., microclimatic constraints) and
inﬂuence the habits of these animals greatly (Altringham
1996), with bats in temperate regions spending up to 20
hours per day in their summer roost (Jenkins et al. 1998).
Consequently, the roost structure is one of the most
important features of a bat’s environment, and the selections
made by bats with respect to the type and location of roost
sites are likely to have a decisive impact on their continued
survival and ﬁtness (Vonhof and Barclay 1996).
Despite prevalent use of anthropogenic structures by bats
and the associated implications for public health, manage-
ment, and bat conservation, little quantitative information
exists on characteristics of roosts in buildings or why bats
select these structures (Kunz and Reynolds 2003). An
exception is microclimate, which has been quantiﬁed as
important in selection of buildings as roosts (Licht and
Leitner 1967, Schowalter and Gunson 1979, Racey and
Swift 1981, Hamilton and Barclay 1994). At least 9 species
of bats in North America commonly roost in buildings, and
many other species occasionally use such structures (Kunz
and Reynolds 2003). Qualitative features of anthropogenic
structures used as bat roosts have been described (see Kunz
1982), but not until recently have studies begun to focus on
measuring the physical attributes of these roost sites
(Entwistle et al. 1997, Williams and Brittingham 1997,
Jenkins et al. 1998). Thus, speciﬁc characteristics that bats
use to select anthropogenic structures at the microhabitat
scale remain largely unknown. Additionally, bats are capable
of moving long distances in a short period of time.
Therefore, multiple scales of investigation may be needed
to determine why they select certain structures (e.g.,
proximity of the roost to water and foraging grounds;
Brigham and Fenton 1986). To our knowledge, no land-
scape-level comparisons between buildings with maternity
roosts and randomly selected buildings have been con-
We selected the big brown bat (Eptesicus fuscus) for our
study because of its well-documented habit of forming
maternity roosts in anthropogenic structures (Barbour and
Davis 1969, Kunz 1982, Agosta 2002) and because of its
importance as an agent of rabies exposure in populated areas
in Colorado and throughout the United States (Pape et al.
1999, Mondul et al. 2003). Despite their known abundance
and proclivity for roosting in buildings during summer,
there is no information beyond anecdotal descriptions of
ecological characteristics for urban roosts of big brown bats
in the western United States. In the eastern United States, a
single study describes detailed characteristics of roost sites of
big brown bats with efforts focused on a speciﬁc type of
728 The Journal of Wildlife Management 71(3)
structure (i.e., houses) in Pennsylvania (Williams and
Historically, big brown bats probably roosted in tree snags
and rock crevices in the eastern foothills and mountains
along Colorado’s Front Range. However, as development
along this plains-mountain interface has expanded so have
the number of anthropogenic roost sites. Subsequently, big
brown bats along the Front Range appear to have made
urban structures their primary maternity-roosting sites
(Fitzgerald et al. 1994). As development along this interface
continues to grow, numbers of big brown bats inhabiting
buildings will likely also grow, increasing the potential for
bat–human interactions. Conﬂicts between bats and humans
in these urban settings may arise from a number of causes,
including bats entering living quarters, noises, odors or
stains caused from guano and urine, ectoparasites, and
potential for disease transmission (Greenhall 1982). Under-
standing which characteristics of anthropogenic roosts these
animals select may assist managers in determining why the
structure was used and how to resolve conﬂicts with owners
of these structures. Thus, the objectives of this study were to
describe maternity roosts selected by big brown bats in an
urbanized setting, to determine the characteristics of these
maternity roosts that differentiate them from randomly
selected buildings that are unoccupied by bats, and to
examine factors inﬂuencing use of buildings as roosts at both
microhabitat and landscape scales.
We conducted summer ﬁeldwork within or on the periphery
of Fort Collins, Colorado. Fort Collins, abutted by the
foothills of the Rocky Mountains to the west and the Cache
La Poudre River to the east, was situated on the northern
end of Colorado’s Front Range. With an elevation of
approximately 1,525 m, the climate consisted of a moderate,
4-season cycle, averaging 300 days of sunshine and 36.8 cm
of precipitation per year (Colorado Climate Center 2002).
Fort Collins experienced its greatest growth in the past half
century, with its population increasing by 30%in the last
decade from 87,758 in 1990 to 126,848 at the beginning of
2003. A similar increase of new buildings occurred, with
new housing permits rising from 774 to 1,703 permits
annually over the last decade. By 2006 the city included
51,236 housing units encompassing 12,950 ha.
During the summers (Jun–Aug) of 2001 to 2004 we
conducted extensive mist-netting for bats in or on the
periphery of Fort Collins along the Poudre River, Spring
Creek, and several parks and open spaces with water sources.
We ﬁtted adult female big brown bats that were pregnant or
lactating with a radiotransmitter and tracked them to
potential maternity colonies. As suggested by Aldridge
and Brigham (1988), we followed a 5%rule of transmitter
mass in relation to body mass, which, for this species, avoids
adverse long-term effects (Neubaum et al. 2005). We
attached radiotransmitters (0.5–0.8 g; Model BD-2AT or
LB-2T, Holohil Systems Ltd., Carp, ON, Canada) to bats
between the scapulae using Skinbond surgical adhesive
(Smith and Nephew United, Inc., Largo, FL). Capture and
handling procedures were approved by animal care and use
committees of Colorado State University (00-074A-01) and
the United States Geological Survey (01-01).
We typically located roosts using radiotelemetry with
some additional roosts located using citizen calls in response
to a newspaper article and ﬂyers. We noted the following
preliminary information for all roosts that we discovered:
address, Universal Transverse Mercator coordinates, build-
ing type, building age, building value, number of stories,
distance to nearest adjacent building, land cover type,
aspects of roost exit(s) and roosting area, nearest obstruc-
tion, and number of roosting and emergence points. We also
took photographs and included a sketched and written
description of the roost.
We categorized some sites as maternity roosts, based on
the sex and number of bats utilizing the structure, the
reproductive condition of the female bats present, and the
time period that the structure was occupied. We used a
combination of evening emergence counts of bats and
capture events, including in-hand captures and evidence of
use based on passive integrated transponder tag reader
technology, to determine if a roost was a maternity colony.
We limited our analysis to maternity roosts that were
utilized by 10 pregnant or lactating female bats and their
pups on a regular basis between 10 June and 20 July. We
derived this maternity period from data collected con-
currently for an associated study (T. J. O’Shea, United States
Geological Survey, unpublished data) based on examination
of 3,360 adult female big brown bats captured in the city
between 2001 and 2004.
Microhabitat characteristics.—To determine roost selectiv-
ity by bats, known roosts must be characterized and
compared with sites that are available but unoccupied
(Entwistle et al. 1997). For each maternity roost we chose a
randomly selected building, unoccupied by bats but
accessible to them, for comparison (Williams and Britting-
ham 1997, Jenkins et al. 1998). We randomly selected
comparison structures using a Geographic Information
System (GIS) coverage of street addresses that was stratiﬁed
by subdivisions so that known roosts could be matched with
a control structure in a similar area of the city. We
thoroughly inspected randomly selected buildings and
substituted them with a new random selection if 1) no
structures were present for the randomly selected address or
if we could not obtain permission or access, 2) we found any
signs of past or present bat use (i.e., guano, urine
crystallization, bat stains, or knowledge of bat use by the
owner), or 3) the building lacked an opening large enough
to allow access to a big brown bat. We measured the same
characteristics at control sites as those at maternity roosts.
We measured characteristics for maternity roosts and
randomly selected buildings at the microhabitat scale. Based
on the literature on bat roosting ecology in anthropogenic
Neubaum et al. Maternity-Roost Selection 729
structures and our own observations, we included 12
variables that provided information about building age,
temperature regimes, and characteristics of exits and interior
roost compartments (Table 1). We determined exit points
for maternity roosts by conducting a thorough survey of the
structure for openings and then conﬁrmed use by conduct-
ing evening emergence counts of bats. Bats used nearly all
openings through which they could ﬁt. Greenhall (1982)
found that big brown bats could use openings 3.2 31.3 cm.
Pilot work from our study found openings with dimensions
as small as 1.0 cm wide or high were used. We only
compared attributes of these buildings with those of
randomly selected buildings that also had openings large
enough for a big brown bat to ﬁt. We considered these
openings exit points in comparisons involving this variable.
If we collected multiple measures for a variable, then we
used the average of the measures for the analysis. We
gathered microclimate data for maternity roosts and
randomly selected buildings using Thermocron iButton
dataloggers (model DS1921, Dallas Semiconductor Corp.,
Dallas, TX), which we placed at the exit points and internal
areas that we determined to be the main roosting locations.
Thus, numbers of dataloggers that we installed varied per
roost depending on accessibility to and usage by the bats.
The main roosting location of maternity roosts was generally
discernable by the presence of guano piles, urine crystal-
lization, and clusters of roosting bats. In randomly selected
roosts where such markers were not available, we used the
locations most similar to those found in maternity roosts.
Dataloggers recorded temperatures (618C) at maternity
roosts and randomly selected buildings bi-hourly through-
out the summer when bats were both present and absent.
We assumed presence of bats did not affect the temperature
of the roosts noticeably due to the large size of the roosting
compartments relative to the bats.
Landscape characteristics.—To assess landscape-scale fea-
tures associated with the locality of the building, we created
a second comparison sample with no stratiﬁcation by
subdivision. Landscape-scale variables included 1) distances
to riparian forest, to rivers and streams, and to roosts
similarly categorized (i.e., used to used, random to random)
and 2) densities of buildings, streets, and trafﬁc counts
(Table 1). We restricted random buildings at this scale to
the total area encompassed by the city and its immediate
periphery. Adult female big brown bats forage during the
summer at distances from ,1.0–18.8 km away from roosts
(Brigham 1991, Everette et al. 2001, Duchamp et al. 2004).
In addition, radiotagged bats in our area utilized feeding
areas over the full length of the town regardless of their
roost location. Therefore, we used the 20 319-km
dimensions of the city and its immediate periphery as the
landscape-scale buffer to randomly select comparison
buildings. We compared habitat features of maternity roosts
and random buildings using previously compiled, spatially
explicit data gathered by others on vegetation structure and
biodiversity (City of Fort Collins 1996). We measured
variables associated with the location of roosts using
distance and density analyses in the ArcInfo and ArcMap
Table 1. Microhabitat- and landscape-scale variables measured at buildings with maternity roosts of big brown bats and randomly selected comparison sites
without roosts in Fort Collins, Colorado, USA, during the summers 2001–2004.
¯SE 95%CI x
Age of building in 2004 (yr)
AOB04 70.0 7.9 53.5–86.5 69.5 7.1 54.6–84.3
Ht of building (m)
HOB 7.9 0.5 6.8–9.0 5.6 0.4 4.8–6.5
No. of exit points
NOEP 1.7 0.2 1.2–2.1 1.1 0.1 0.9–1.3
Ht of exit point from ground (m) HEPFG 5.7 0.5 4.8–6.8 4.0 0.3 3.4–4.7
Compartments in roosting area
NOCIRA 2.3 0.4 1.5–3.0 1.2 0.1 1.0–1.4
Dimension of exit point: ht (cm) DOEPH 14.1 6.0 1.6–26.7 3.8 0.7 2.4–5.1
Dimension of exit point: width (cm) DOEPW 36.9 9.0 18.2–55.5 10.0 3.4 3.0–17.0
Area of exit point (cm
) AREA 164.8 46.4 68.4–261.2 21.8 4.4 12.8–30.9
Ht of roost compartment (m)
DORCH 1.5 0.3 0.9–2.1 1.2 0.2 0.7–1.7
Nearest obstruction to exit point (m)
NOTED 5.3 0.6 4.1–6.5 3.9 1.1 1.6–6.2
Temp of roosting area (8C)
TAVG 24.8 0.6 23.6–26.0 24.4 0.5 23.4–25.5
Daily roost temp variation (8C)
TSE 1.6 0.1 1.3–1.9 1.9 0.2 1.6–2.3
Distance to riparian forest (m)
DISTRF 550 51 448–653 580 57 465–694
Distance to rivers and streams (m)
DISTRS 1,080 110 860–1,302 1,475 127 1,220–1,730
Distance to all similarly categorized roosts (km) DTASCR 220 10 200–241 320 12 295–345
Building density (No./km) BDL 1,584 141 1,302–1,867 1,689 123 1,443–1,935
Street density (No./km) SDL 18.2 1.2 15.7–20.7 15.4 0.9 13.7–17.2
Traffic count density (No./km)
TCDL 104,911 13,125 78,535–131,287 65,319 8,368 48,504–82,134
n¼22 and 50 roost sites for the microhabitat- and landscape-scale analyses, respectively, and an equal no. of randomly selected comparison sites.
Williams and Brittingham 1997.
Entwistle et al. 1997.
Jenkins et al. 1998.
Hamilton and Barclay 1994, Licht and Leitner 1967.
Gehrt and Chelsvig 2004.
730 The Journal of Wildlife Management 71(3)
applications of ArcGIS (version 8.2). We collected straight-
line distances to water and likely foraging grounds (e.g.,
riparian forest, rivers and streams) and to other similarly
categorized roosts using the Join by spatial location function
in ArcMap and the POINTDISTANCE command in
ArcInfo. The distance to other similarly categorized roosts
variable was a total measure of the distance from one roost
to all other roosts of the same category (i.e., used to used, or
random to random) and provided a measure of spatial
aggregation for each roost. To determine whether patterns
of urbanization were correlated with roost-site location, we
measured density of buildings, streets, and trafﬁc around
roost sites. We used the Animal Movement extension for
ArcView (Hooge and Eichenlaub 1997) and least-squares
cross-validation (Silverman 1986, Worton 1989) to deter-
mine a search radius of approximately 1 km for use in the
Model Development and Analyses
We pooled data from 2001 to 2004 for analyses to compare
maternity roosts and randomly selected buildings. Pooling
data to increase sample size (which allowed for analysis of
selection rather than description only) is justiﬁed in
situations where few observations occur for many radio-
tagged animals (White and Garrott 1990). In addition, all
maternity roosts that we considered in this study were used
by bats in repeated years, so effects that could be masked by
pooling (e.g., yr) seemed unlikely. We summarized
characteristics of maternity roosts (x¯, SE, CI) and compared
them with those characteristics of randomly selected
buildings at the microhabitat and landscape scales using
SAS (SAS Institute, Inc., Cary, NC).
We modeled roost-site selection, including physical
characteristics (e.g., No. of emergence points, building
age, entrance dimensions) at the microhabitat scale using
logistic regression (Proc GENMOD) with site type (known
maternity roost vs. randomly selected building) as the
response variable and site characteristics as the independent
variables (Table 1). We quantiﬁed landscape-scale variables
using GIS and modeled them using logistic regression. We
considered our logistic regression analyses to be a case-
control design (Keating and Cherry 2004), because we
searched all randomly selected buildings and veriﬁed their
We based the variables used in our linear models on a
literature review of roosting ecology for big brown bats and
our own observations made in 2001–2002. To eliminate
variable redundancy (Burnham and Anderson 2002), we
dropped some variables that were initially considered
because they were highly correlated with other variables
(Proc CORR). We compared models with all possible
combinations of the 12 independent variables at the
microhabitat scale and 6 variables at the landscape scale
using logistic regression. We chose this exploratory
approach (see Burnham and Anderson 2002) because very
little speciﬁc information exists from past research for
development of a priori models. Further, we felt that the
various combinations of variables in the models were
biologically meaningful based on our current knowledge of
big brown bat ecology in urban settings. We focused on
which variables, rather than models, were most important by
Burnham and Anderson 2002). Multimodel inference using
importance values requires that a balanced model set (i.e., all
variables appear in an equal number of models) be
considered (Burnham and Anderson 2002), and it is helpful
in situations where a priori predictions are difﬁcult to state
due to a lack of background information.
Models derived from the data were ranked using Akaike’s
Information Criterion corrected for sample size ([AIC
Burnham and Anderson 2002). We calculated Akaike
; probability that the ith model is actually the
best approximating model among the candidate models) and
relative importance (w
) for each predictor variable by
summing Akaike weights across all models in the set
(Burnham and Anderson 2002). We also report the
unconditional parameter estimates (~
b) and associated stand-
ard errors for each variable, based on model averaging over
all models in the set (Burnham and Anderson 2002).
During the summers of 2001 to 2004 we located 142 roosts
used by big brown bats. Fifty of these roosts met the criteria
for maternity roosts, and we used 33 in microhabitat-scale
analysis. The remaining 17 roosts could not be measured at
this scale due to various logistical constraints (primarily
owner permission and accessibility). In addition, we did not
assess temperature in 11 of the 33 maternity roosts because
of lack of internal access and limited numbers of temper-
ature loggers. Thus, we used a ﬁnal subset of 22 known
maternity roosts and their associated 22 randomly selected
comparison buildings for the microhabitat analysis (total n¼
44). This sample included a variety of different structures:
21 houses, 3 schools, 8 businesses, 4 apartments, 2 churches,
2 garages, 2 barns, and 2 bat boxes that were afﬁxed to the
sides of buildings. Because we used GIS to conduct the
analysis at the landscape scale, the constraints encountered
at the microhabitat scale were not a factor. As a result, we
considered the full sample of 50 roosts that met the
maternity-roost criteria along with 50 random buildings in
the landscape scale analysis (total n¼100).
Microhabitat and Landscape Characteristics
Using the 12 independent variables (Table 1), all possible
model combinations resulted in 4,096 models including an
intercept-only model. Consequently, the relative weights
) for the top models (i.e., within 2 AIC
relatively small (i.e., low probability of support for any one
model; Table 2), but our focus was on relative importance
) of these variables in predicting maternity-roost
site use and not on rankings of the various models. The area
of the exit point variable (AREA; ~
b¼0.069, SE ¼0.001)
had the greatest relative importance (w
twice as large as the next closest variable (Table 3), was
present in all top models (Table 2), and appeared in the top
506 models. Known maternity roosts, on average, had
Neubaum et al. Maternity-Roost Selection 731
openings covering almost 8 times as much area compared to
comparison buildings, with nonoverlapping 95%conﬁdence
intervals (Table 1). Opening width (DOEPW; ~
SE ¼0.003) and height (DOEPH; ~
b¼0.007, SE ¼0.005)
of exit points, which we used to calculate AREA, had much
lower relative importance values (w
¼0.40 and 0.22,
respectively; Table 3) with width almost twice as important
as height. Average roost temperature (TAVG; ~
SE ¼0.069) and exit height from the ground (HEPFG; ~
0.245, SE ¼0.136) were the second and third most
important variables (w
¼0.50 and 0.48, respectively; Table
3). Both TAVG and HEPFG appeared in 8 of the top 15
models. TAVG from maternity roosts were warmer than
randomly selected buildings but with greatly overlapping
95%conﬁdence intervals; known roosts had exit points that
were on average 1.7 m higher than comparable buildings
with nonoverlapping 95%conﬁdence intervals (Table 2).
Height of the building (HOB; ~
b¼0.179, SE ¼0.089)
appears in 7 of the top models and had the fourth highest
importance value (w
¼0.44). Maternity roosts were on
average 2.3 m higher than randomly selected buildings. The
remaining variables were assigned importance values of
0.40, indicating that in comparison to the top variable
(area of the exit point) they carried considerably less
importance in inﬂuencing maternity-roost site selection by
big brown bats.
We also conducted a post hoc analysis based on our
knowledge gained from the exploratory analysis to consider
several interaction variables. We considered 16 exploratory
models that had a difference in AIC
score between the ith
and top-ranked model (D
),2 in the original analysis and 7
new models in our post hoc analysis. The new models
included 2 interaction variables: 1) average temperature 3
area of the exit point and 2) average temperature 3height of
the exit point from the ground; we modeled both interaction
variables alone, together, and with several variables from the
exploratory analysis. The top model (AIC
0.20 based on the 23 models in the post hoc analysis)
included the dimension of exit point – width variable and
the 2 interaction variables (x¯ temp 3area, x¯ temp 3ht of the
exit point from the ground). The top 4 ranked models
(range of D
¼0–2.41) included the average temperature 3
area interaction, and the top 2 models included the average
temperature 3height of the exit point from the ground
interaction. Both interaction variables had positive slopes (~
¼0.004, SE ¼0.001 and ~
b¼0.027, SE ¼0.015,
respectively) indicating more likely use as maternity roosts.
Thus, used roosts provide circumstances where areas of the
Table 2. Rankings by Akaike’s Information Criterion adjusted for small sample size (AIC
; Burnham and Anderson 2002) of top logistic regression models
(Proc GENMOD [SAS Institute, Inc., Cary, NC]) comparing structures used by big brown bats as maternity roosts to randomly selected comparison sites in
Fort Collins, Colorado, USA, from 2001–2004. D
¼difference in AIC
score between the ith and top-ranked model; w
¼Akaike weight. We modeled 12
variables at the microhabitat scale and 6 variables at the landscape scale; we define variables in Table 1.
Scale Model KAIC
Microhabitat HEPFG þDOEPW þAREA 5 37.34 0.00 0.016
NOEP þHEPFG þTAVG þAREA 6 37.61 0.28 0.014
HOB þDOEPW þAREA 5 37.75 0.42 0.013
HEPFG þDOEPW þTAVG þAREA 6 37.87 0.53 0.012
HOB þNOEP þTAVG þAREA 6 37.89 0.55 0.012
HOB þAREA 4 38.22 0.88 0.010
HEPFG þTAVG þAREA 5 38.59 1.25 0.009
HOB þNOEP þAREA 5 38.61 1.27 0.009
HEPFG þAREA 4 38.73 1.40 0.008
HOB þTAVG þAREA 5 38.97 1.63 0.007
HEPFG þDORCH þAREA 5 39.00 1.66 0.007
HEPFG þTAVG þTSE þAREA 6 39.01 1.67 0.007
HEPFG þDOEPW þDORCH þAREA 6 39.02 1.68 0.007
HOB þTAVG þTSE þAREA 6 39.04 1.71 0.007
HOB þDOEPW þTAVG þAREA 6 39.26 1.92 0.006
Landscape DTASCR þBDL þSDL þTCDL 6 93.44 0.00 0.290
DISTRS þDTASCR þBDL þSDL þTCDL 7 93.82 0.39 0.239
DISTRF þDTASCR þBDL þSDL þTCDL 7 95.00 1.57 0.133
Table 3. Relative variable importance (w
) of variables (Burnham and
Anderson 2002) in predicting use of structures as maternity roosts by big
brown bats in Fort Collins, Colorado, USA, from 2001 to 2004 based on
logistic regression analysis at microhabitat and landscape scales. We define
variables in Table 1.
Scale Variable w
Microhabitat AREA 0.99
Landscape DTASCR 1.00
732 The Journal of Wildlife Management 71(3)
exit points are larger, and the roosting area maintains
slightly warmer temperatures despite having the potential
for temperature ﬂuctuations due to easier exchange of
At the landscape scale, the 6 independent variables (Table
1) resulted in 64 models (including the intercept-only
model) being considered (Table 2). Distance to all similarly
categorized maternity roosts (DTASCR; ~
SE ¼0.004 310
) and building density (BDL; ~
SE ¼0.001 310
) appeared in the top 32 and 16 models,
respectively, and were both assigned the highest importance
¼1.00 (Table 3). The DTASCR for maternity
roosts was 220 km versus 320 km for randomly selected
comparison buildings (distances were the sum of all similarly
categorized roosts) with conﬁdence intervals that did not
overlap. The BDL showed a similar trend, with used roosts
having 105 fewer buildings/km
but with conﬁdence
intervals that widely overlapped those for randomly selected
buildings (Table 1). Thus, the negative estimates for
DTASCR and BDL indicate that known maternity roosts
on the landscape level were more closely spaced to other
known maternity roosts occurred in areas with lower
building densities than randomly selected buildings. The 2
remaining urbanization variables, street and trafﬁc density
b¼0.266, SE ¼0.033; TCDL, ~
SE ¼0.007 310
), were also considered important with w
¼0.84 and 0.81, respectively. These parameter estimates
suggested that known maternity roosts were surrounded by
more streets with lower trafﬁc volumes than randomly
selected buildings. The distance to rivers and streams
variable (DISTRS; ~
and the distance to riparian forest variable (DISTRF; ~
) had smaller importance
¼0.42 and 0.29, respectively) and occurred much
less frequently in the top models. Both of these variables had
negative slopes suggesting that known maternity roosts were
situated closer to water and riparian forest in the landscape
than random buildings.
In Fort Collins, Colorado, a variety of anthropogenic
structures with similar key characteristics were selected by
big brown bats as maternity roosts. We found that the area
of the exit point was the most important variable in
distinguishing structures with known roosts from randomly
selected structures without roosts. Larger exits may increase
access to the roost. Ease of access may be particularly
important to maternity colonies where larger exits may
facilitate ingress and egress rates. Williams and Brittingham
(1997) noted that big brown bats in Pennsylvania used
buildings that were more accessible based on numbers of
exits but suggested that additional factors might be more
important in inﬂuencing roost selection. Our results suggest
that area of the exit point was more important than number
of exit points. Our ﬁndings also imply that the interaction of
width and height is more important than these variables
alone. Roosts with larger openings offer access by more than
one individual at a time, and this may be more important as
juvenile bats mature and begin exiting the roost. Juvenile
bats inexperienced in ﬂying may take longer to exit from an
access point and therefore block other bats attempting to
enter or exit.
Several studies have found temperature of anthropogenic
roosts to be important in summer because warmer
conditions have positive effects on reproduction and
development of young (Schowalter and Gunson 1979,
Racey and Swift 1981) and on energy conservation
(Hamilton and Barclay 1994, Lausen and Barclay 2003).
However, our known roosts were only slightly warmer than
random buildings (Table 1). This small difference may
suggest that bats are sensitive to minute microclimatic
differences in the roosting environment. However, it is more
likely that randomly selected buildings provided micro-
climates within a range of suitability for bats but were not
selected because they did not occur in combination with
other important variables (e.g., lacked suitable openings for
access). Randomly selected buildings had slightly wider daily
temperature ﬂuctuations than known structures, but this
variable was not considered important compared to the
other variables examined. This also underscores the
exploratory nature of our research; future research might
focus on studying the interaction effects of variables (e.g.,
exit area and temp), as suggested by our post hoc analysis
The height of the exit point from the ground variable was
also important in maternity-roost selection by big brown
bats, as it was when selecting hibernacula (Neubaum et al.
2006). Maternity roosts with higher exit points may provide
bats, especially young individuals inexperienced in ﬂying,
with added room to drop and obtain lift, thus avoiding low-
lying clutter or potential urban ground predators. Our post
hoc analysis indicated that average temperature of the
roosting area could also be affected by the height of the exit
from the ground: higher roosts may get more solar radiation
because they are shaded less by trees and adjacent buildings.
Future research should investigate how these and other
variables interact to affect roosting environments. Similarly,
buildings used by maternity colonies were notably higher
than randomly selected buildings. This added height might
provide bats with a dominant object to perceive and cue on,
making the roost easier to locate upon returning from
The landscape variable with the greatest relative impor-
tance was distance to all similarly categorized roosts. Known
maternity roosts were located closer to other known roosts
in Fort Collins than random buildings were located to each
other. Because bats are often evicted from urban roosts for a
variety of reasons (see Kunz and Reynolds 2003), nearby
alternative roosts would decrease time and energy spent on
relocating to a new roost. Lewis (1995) found that roost-site
ﬁdelity for bats varied depending on the availability and
permanency of the roosts used and suggested that ﬁdelity
was high at buildings for these reasons. However, despite
Neubaum et al. Maternity-Roost Selection 733
the relative permanency of the anthropogenic roost
structures in Fort Collins, availability of these roosts
sometimes changed during our study, with bats eventually
being excluded from 30%of used maternity roosts because
of owner intolerance or building improvements. Similarly, in
more natural settings, maternity roosts of big brown bats in
trees have been situated closer to other usable trees (Vonhof
and Barclay 1996, Rabe et al. 1998, Cryan et al. 2001).
Willis and Brigham (2004) note that the ﬁssion–fusion
social organization of big brown bats includes frequent
roost-switching, which is important in the maintenance of
social relationships. They suggest that use of multiple roosts
in trees, caves, mines and buildings may be equivalent
reﬂections of this social system. In addition, the area of Fort
Collins with the highest density of roosts is located where
the original settlement began. Conceivably, bats began to
utilize structures in this area ﬁrst before splitting off to form
new colonies, resulting in greater roost densities in this area.
Thus, having multiple roosts may be a carry-over behavior
that originated in the historical roosting habits of these bats
in tree snags and rock crevices.
The building density variable also had a high relative
importance value, only slightly below that of the distance to
all maternity roosts, with known roosts negatively correlated
with building density compared to randomly selected
buildings. These ﬁndings suggest there may be an urban-
ization tolerance level for bats. However, a bias in the
building-density measurement was likely, because a large
proportion of the known maternity roosts occurred in the
older subdivisions of the city. Newer developments are
encouraged to construct buildings at higher densities (i.e.,
smaller lot sizes relative to house size). The older
subdivisions of the city incorporate a variety of different
structures, especially in commercial areas, and tend to have
larger footprints and encompass more area in comparison to
smaller buildings in housing tracts. Thus, fewer buildings
would have been counted in the density analysis for the areas
surrounding known roosts because larger buildings were
accounting for more of the space. Further, several randomly
selected buildings were situated near trailer parks, which
may have inﬂated building densities. Thus, in future studies,
considering the areas of building footprints rather than
numbers of buildings for building density may provide a
better measure of this urbanization variable.
Barbour and Davis (1969) suggest that big brown bats in
North America have adapted so well to anthropogenic
structures during maternity seasons that records for natural
roosts have become rare. Interestingly, the most important
variables in our study of urban roosts were similar to studies
that have characterized selection of natural maternity roosts
(i.e., tree snags and rock crevices) of big brown bats. Tree
snags used as maternity roosts are taller than random trees
(Vonhof and Barclay 1996, Rabe et al. 1998, Cryan et al.
2001) and are available in higher densities in the surround-
ing area. Kalcounis and Brigham (1998) found wider
openings and warmer microclimates for secondary cavities
used by big brown bats in aspen (Populus tremuloides) trees.
Lausen and Barclay (2002, 2003) found rock crevices used as
maternity roosts by big brown bats were higher above the
ground, had wider openings, and had microclimates that
were stable and warmer at night when mothers would be
absent. Urban roosting options in Fort Collins appear to
have similar characteristics and seem to be plentiful,
allowing bats to be highly selective.
In addition, food resources on the Front Range may be
more abundant than in the adjacent foothills. Nonnative
deciduous trees, agricultural ﬁelds and open spaces may
provide a variety of preferential foods for big brown bats
(Whitaker 1995). The Poudre River corridor running the
length of the city, along with canals, drainage ditches, and
water impoundments, is surrounded by riparian forest,
which provides a highly altered landscape that was
historically short-grass prairie (Knopf 1986). Such riparian
habitat likely has provided dramatic increases in drinking
and foraging areas along the Front Range available to these
bats. As a result of plentiful roosting and feeding
opportunities, big brown bats may be enjoying high
reproductive success relative to natural settings, which, in
turn, suggests that their populations have increased. Others
also suggest that big brown bat populations have increased
in numbers (Whitaker and Gummer 1992, 2000) and
geographic distribution (Fitzgerald et al. 1994, Whitaker
and Gummer 2000) due to the increased presence of
anthropogenic structures over the past century.
In Colorado, big brown bats were the most common species
submitted to local health departments for rabies testing and
the most common bat to bite humans (Pape et al. 1999).
Such interactions between humans and bats are often
associated with a negative outcome for both species.
Therefore, understanding and managing the types of
anthropogenic roosts that these animals prefer is a ﬁrst step
in predicting and preventing such conﬂicts. Alternatives to
extermination, such as decreasing the access point size,
alteration of the microclimate in the roosting area, or
installation of an object that reduces the height of the exit
point and interferes with the drop of the ﬂight path all may
deter use of a building by bats. Such modiﬁcations can be
combined with humane exclusion practices during seasons
when bats are absent (Greenhall 1982). Provision of well-
designed bat boxes (Brittingham and Williams 2000),
installed near the original exit and following speciﬁcations
based on important variables found in this study should be
examined for effectiveness in providing an alternative to bats
other than shifting to another building. Big brown bats
show high ﬁdelity to urban roosts and will return if
exclusion attempts are not thorough (Brigham and Fenton
1986, Lewis 1995). Because big brown bats appear to
selectively choose speciﬁc combinations of characteristics
found at maternity roosts, not all structures that are available
can be considered suitable by managers. Therefore, exclud-
ing bats from well-established maternity roosts could have
negative impacts on bat populations (Brigham and Fenton
734 The Journal of Wildlife Management 71(3)
1986) and may be undesirable from the standpoint of bat
conservation. In such cases managers should consider
alternative remedial practices.
We thank T. Barnes, J. Boland, M. Carson, K. Castle, P.
Cryan, L. Ellison, D. Emptage, L. Galvin, D. Grossblat, M.
Hayes, B. Iannone, J. LaPlante, H. Lookingbill, G. Nance,
M. Neubaum, R. Pearce, V. Price, S. Smith, J. Tharp, and
T. Torcoletti for their assistance in the ﬁeld. We are greatly
appreciative of R. Bowen, J. Wimsatt, and L. Ansell for
their technical and data management support, R. Coleman,
T. Fancher, T. Giles, M. Laituri, and B. Waltermire for
assistance with GIS tasks, D. Anderson, P. Doherty, and T.
Stanley for statistical insight, and R. Reich and J. Savidge
for helpful comments on earlier drafts. We are also grateful
to all of the property owners that agreed to facilitate this
study by allowing access to their buildings. This research
was supported by the United States Geological Survey and a
grant from the National Science Foundation and National
Institute of Health Ecology of Infectious Disease Program
(0094959) to Colorado State University.
Agosta, S. J. 2002. Habitat use, diet and roost selection by the big brown
bat (Eptesicus fuscus) in North America: a case for conserving an abundant
species. Mammal Review 32:179–198.
Aldridge, H. D. J. N., and R. M. Brigham. 1988. Load carrying and
maneuverability in an insectivorous bat: a test of the 5%‘‘Rule’’ of
radiotelemetry. Journal of Mammalogy 69:379–382.
Altringham, J. D. 1996. Bats: biology and behaviour. Oxford University
Press, New York, New York, USA.
Barbour, R. W., and W. H. Davis. 1969. Bats of America. University Press
of Kentucky, Lexington, USA.
Brigham, R. M. 1991. Flexibility in foraging and roosting behaviour by the
big brown bat (Eptesicus fuscus). Canadian Journal of Zoology 69:117–
Brigham, R. M., and M. B. Fenton. 1986. The influence of roost closure on
the roosting and foraging behaviour of Eptesicus fuscus (Chiroptera:
Vespertilionidae). Canadian Journal of Zoology 64:1128–1133.
Brittingham, M. C., and L. M. Williams. 2000. Bat boxes as alternative
roosts for displaced bat maternity colonies. Wildlife Society Bulletin 28:
Burnham, K. P., and D. R. Anderson. 2002. Model selection and
multimodel inference: a practical information-theoretic approach. Second
edition. Springer-Verlag, New York, New York, USA.
City of Fort Collins. 1996. Geographic Information Services Division. City
of Fort Collins, Fort Collins, Colorado, USA. ,http://www.ci.
fort-collins.co.us/gis/.. Accessed 22 Apr 2003.
Colorado Climate Center. 2002. Department of Atmospheric Science.
Colorado State University, Fort Collins, Colorado, USA. ,http://ccc.
atmos.colostate.edu/.. Accessed 1 Dec 2004.
Cryan, P. M., M. A. Bogan, and G. M. Yanega. 2001. Roosting habits of
four bat species in the Black Hills of South Dakota. Acta Chiropter-
Duchamp, J. E., D. W. Sparks, and J. O. Whitaker, Jr. 2004. Foraging-
habitat selection by bats at an urban–rural interface: comparison between
a successful and a less successful species. Canadian Journal of Zoology 82:
Entwistle, A. C., P. A. Racey, and J. R. Speakman. 1997. Roost selection by
the brown long-eared bat Plecotus auritus. Journal of Applied Ecology 34:
Everette, A. L., T. J. O’Shea, L. E. Ellison, L. A. Stone, and J. L.
McCance. 2001. Bat use of a high plains urban wildlife refuge. Wildlife
Society Bulletin 29:967–973.
Fitzgerald, J. P., C. A. Meaney, and D. M. Armstrong. 1994. Mammals of
Colorado. Denver Museum of Natural History and University Press of
Colorado, Niwot, USA.
Gehrt, S. D., and J. E. Chelsvig. 2004. Species-specific patterns of bat
activity in an urban landscape. Ecological Applications 14:625–635.
Greenhall, A. M. 1982. House bat management. U.S. Fish and Wildlife
Service Report, Resource Publication 143, Washington, D.C., USA.
Hamilton, I. M., and R. M. R. Barclay. 1994. Patterns of daily torpor and
day-roost selection by male and female big brown bats (Eptesicus fuscus).
Canadian Journal of Zoology 72:744–749.
Hooge, P. N., and B. Eichenlaub. 1997. Animal movement extension to
Arcview, version 1.1. U.S. Geological Survey, Alaska Science Center–
Science Office, Anchorage, USA.
Jenkins, E. V., T. Laine, S. E. Morgan, K. R. Cole, and J. R. Speakman.
1998. Roost selection in the pipistrelle bat, Pipistrellus pipistrellus
(Chiroptera: Vespertilionidae), in northeast Scotland. Animal Behaviour
Kalcounis, M. C., and R. M. Brigham. 1998. Secondary use of aspen
cavities by tree-roosting big brown bats. Journal of Wildlife Management
Keating, K. A., and S. Cherry. 2004. Use and interpretation of logistic
regression in habitat-selection studies. Journal of Wildlife Management
Knopf, F. L. 1986. Changing landscapes and the cosmopolitism of the
eastern Colorado avifauna. Wildlife Society Bulletin 14:132–142.
Kunz, T. H. 1982. Roosting ecology of bats. Pages 1–55 in T. H. Kunz,
editor. Ecology of bats. Plenum Press, New York, New York, USA.
Kunz, T. H., and D. S. Reynolds. 2003. Bat colonies in buildings. Pages
91–102 in T. J. O’Shea and M. A. Bogan, editors. Monitoring trends in
bat populations of the United States and Territories: problems and
prospects. U.S. Geological Survey, Biological Resources Discipline,
Information and Technology Report, USGS/BRD/ITR-2003-0003,
Washington, D.C., USA.
Lausen, C. L., and R. M. R. Barclay. 2002. Roosting behavior and roost
selection of female big brown bats (Eptesicus fuscus) roosting in rock
crevices in southeastern Alberta. Canadian Journal of Zoology 80:1069–
Lausen, C. L., and R. M. R. Barclay. 2003. Thermoregulation and roost
selection by reproductive female big brown bats (Eptesicus fuscus) roosting
in rock crevices. Journal of Zoology 260:235–244.
Lewis, S. E. 1995. Roost fidelity of bats: a review. Journal of Mammalogy
Licht, P., and P. Leitner. 1967. Behavioral responses to high temperatures
in three species of California bats. Journal of Mammalogy 48:52–61.
Mondul, A. M., J. W. Krebs, and J. E. Childs. 2003. Trends in national
surveillance for rabies among bats in the United States (1993–2000).
Journal of American Veterinary Medical Association 222:633–639.
Neubaum, D. J., M. A. Neubaum, L. E. Ellison, and T. J. O’Shea. 2005.
Survival and condition of big brown bats (Eptesicus fuscus)after
radiotagging. Journal of Mammalogy 86:95–98.
Neubaum, D. J., T. J. O’Shea, and K. R. Wilson. 2006. Autumn migration
and selection of rock crevices as hibernacula by big brown bats in
Colorado. Journal of Mammalogy 87:470–479.
Pape, W. J., T. D. Fitzsimmons, and R. E. Hoffman. 1999. Risk for rabies
transmission from encounters with bats, Colorado, 1977–1996. Emerging
Infectious Diseases 5:433–437.
Pierson, E. D. 1998. Tall trees, deep holes, and scarred landscapes. Pages
309–325 in T. H. Kunz and P. A. Racey, editors. Bat biology and
conservation. Smithsonian Institution Press, Washington, D.C., USA.
Rabe, M. J., T. E. Morrell, H. Green, J. C. deVos, Jr., and C. R. Miller.
1998. Characteristics of ponderosa pine snag roosts used by reproductive
bats in northern Arizona. Journal of Wildlife Management 62:612–621.
Racey, P. A., and S. M. Swift. 1981. Variations in gestation length in a
colony of pipistrelle bats (Pipistrellus pipistrellus) from year to year. Journal
of Reproduction and Fertility 61:123–129.
Schowalter, D. B., and J. R. Gunson. 1979. Reproductive biology of the big
brown bat (Eptesicus fuscus) in Alberta. The Canadian Field-Naturalist 93:
Silverman, B. W. 1986. Density estimation for statistics and data analysis.
Chapman and Hall, New York, New York, USA.
Vonhof, M. J., and R. M. R. Barclay. 1996. Roost-site selection and
Neubaum et al. Maternity-Roost Selection 735
roosting ecology of forest-dwelling bats in southern British Columbia.
Canadian Journal of Zoology 74:1797–1805.
Whitaker, J. O., Jr. 1995 Food of the big brown bat Eptesicus fuscus from
maternity colonies in Indiana and Illinois. American Midland Naturalist
Whitaker, J. O., Jr., and S. L. Gummer. 1992. Hibernation of the big
brown bat, Eptesicus fuscus, in buildings. Journal of Mammalogy 73:312–
Whitaker, J. O., Jr., and S. L. Gummer. 2000. Population structure and
dynamics of big brown bats (Eptesicus fuscus) hibernating in buildings in
Indiana. American Midland Naturalist 143:389–396.
White, G. C., and R. A. Garrott. 1990. Analysis of wildlife radiotracking
data. Academic Press, San Diego, California, USA.
Williams, L. M., and M. C. Brittingham. 1997. Selection of maternity
roosts by big brown bats. Journal of Wildlife Management 61:359–368.
Willis, C. K. R., and R. M. Brigham. 2004. Roost switching, roost sharing
and social cohesion: forest-dwelling big brown bats, Eptesicus fuscus,
conform to the fission–fusion model. Animal Behaviour 68:495–505.
Worton, B. J. 1989. Kernel methods for estimating the utilization
distribution in home-range studies. Ecology 70:164–168.
Associate Editor: Chambers.
736 The Journal of Wildlife Management 71(3)