Tools and Technology
Efﬁcacy of Scent-Detection Dogs for Locating
Bat Roosts in Trees and Snags
CAROL L. CHAMBERS,
School of Forestry, Northern Arizona University, Flagstaff, AZ 86011, USA
CHRISTINA D. VOJTA, Landscape Conservation Initiative, Northern Arizona University, Flagstaff, AZ 86011, USA
ELISABETH D. MERING, School of Forestry, Northern Arizona University, Flagstaff, AZ 86011, USA
BARBARA DAVENPORT, PackLeader Dog Training LLC, 14401 Crews Road KPN, Gig Harbor, WA 98329, USA
ABSTRACT Conservation efforts for tree- and snag-roosting bats are challenging because roost sites are
difﬁcult and costly to locate. We assessed the ability of scent detection dogs to locate bat roosts using
controlled ﬁeld experiments conducted at 2 sites in northern Arizona, USA, during July and August of 2007,
with small bags of bat guano and known bat roosts. Scent detection dogs correctly detected 79% of guano bags
and 29% of known bat roosts; but, when searching during shorter time periods for roosts only, dogs correctly
detected 77% of known roosts. Factors affecting detection of guano bags included bag height and mass.
Factors affecting detection of roosts included height, size of the bat colony, and air temperature. Rest breaks
and competition between dogs apparently increased their success rate in ﬁnding roosts. Using a scent
detection dog to locate roosts was similar in cost to radiotelemetry; however, the use of a scent detection dog
was less invasive because the use of dogs precluded the need to capture and radiotag bats. Important
considerations inﬂuencing the efﬁcacy of using scent detection dogs include training and transportation costs,
and skill level of dog handlers. Although some factors limit the ability of scent detection dogs to accurately
locate roosts, dogs in our study approximated the roost location to within a 30-m radius, which may be
sufﬁcient to protect these areas during management activities. If managers have the ability within an area to
retain all large-diameter snags that have bat roost characteristics, then scent detection dogs will likely not
provide adequate added value to warrant their survey costs. However, scent dogs could be important for
locating roosts of species that are imperiled or of conservation concern. Ó2015 The Wildlife Society.
KEY WORDS Arizona, Chiroptera, feces, field trials, forest, guano, Idionycteris phyllotis, radiotelemetry, roost.
Most conservation efforts for bats focus on conspicuous
species that aggregate in large numbers, are impacted by focal
threats (e.g., impacts to cave and mine roosts), and are easy to
observe (Weller et al. 2009). In contrast, inconspicuous
species that roost in small, dispersed colonies have received
fewer conservation actions because colonies are hard to locate
and the effects on these species from diffuse threats (e.g.,
changes in forest structure or land conversion) are difﬁcult to
document (Weller et al. 2009). Moreover, species that roost
in crevices, snags, and trees tend to switch roost sites
frequently during the breeding season (e.g., Bernardos et al.
2004, Solvesky and Chambers 2009); this behavior creates
additional challenges for ﬁnding and protecting roosts.
Since the 1980s, resource managers and researchers have
used radiotelemetry to locate roosts, and the number of
telemetry studies has increased with the development of
micro-transmitters (Amelon et al. 2009). Nevertheless,
radiotelemetry presents challenges because the life of
micro-transmitters is generally <2 weeks (depending on
the size of the transmitter, which is conditioned by the mass
of the bat; Aldridge and Brigham 1988), and the task of
locating telemetered bats is labor-intensive and therefore
expensive. Moreover, bats can roost substantial distances
from their original points of capture (e.g., Corbett et al.
2008, Chambers et al. 2011), so roosts located through
radiotelemetry are often outside speciﬁc areas targeted for
management. In these situations, it is difﬁcult to justify the
telemetry costs when there are no obvious beneﬁts to
Scent detection dogs have successfully located live animals
(Vice and Engeman 2000, Nussear et al. 2008, Gsell et al.
2010), carcasses (Arnett 2006, Paula et al. 2011, Mathews
et al. 2013), and scat for species including North Atlantic
right whales (Eubalaena glacialis), bobcat (Lynx rufus), bears
(Ursus arctos and U. americanus), and bush dogs (Speothos
venaticus; Wasser et al. 2004, Harrison 2006, Rolland et al.
2006, Long et al. 2007b, DeMatteo et al. 2009). If scent
detection dogs (hereafter, dogs) could reliably locate bat
roosts, studies could target bat activity within speciﬁc areas
affected by land management or disturbance (e.g., timber
sales, housing developments, ﬁres, mining) instead of an
Received: 25 August 2014; Accepted: 29 June 2015
Published: 22 October 2015
Wildlife Society Bulletin 39(4):780–787; 2015; DOI: 10.1002/wsb.598
780 Wildlife Society Bulletin 39(4)
emphasis on locations of telemetered individuals, which may
be outside areas of interest.
Early experiments showed that dogs could be trained to use
bat guano to locate roosts in short-statured woodlands of
pinyon pine (Pinus edulis) and juniper (Juniperus spp.;
A. Chung-McCoubrey, National Park Service, unpublished
data). Although these results were encouraging, the experi-
ments were not designed to evaluate use of dogs in forested
environments or quantify rates of success under speciﬁc
environmental conditions. To provide a quantitative assess-
ment of the efﬁcacy of using dogs, we developed a set of
controlled ﬁeld experiments using different dog–handler pairs
under measured environmental conditions, and with con-
trolled levels of guano and known roost sites. Our speciﬁc
objectives were to identify external factors affecting roost
detection (e.g., dog–handler pair, weather, size and height of
bat roost), generate cost estimates for using dogs as an
alternative to radiotelemetry, and develop recommendations
for using dogs in area-based studies of management activities.
We conducted the study at 2 sites on the Coconino National
Forest in northern Arizona, USA, during July and Au-
gust 2007: Fort Valley (358170N, 1118430W, elevation 2,160–
2,790 m) and Happy Jack (348400N, 1118230W, elevation
2,050–2,290 m). Slopes rangefrom 0% to 29%. Ponderosa pine
(Pinus ponderosa) dominated the areas, but other tree species
include Gambel oak (Quercus gambelii) below 2,250 m and
mixed conifer (Abiesand Picea spp.) above 2,400 m. Understory
vegetation was generally sparse and open. During summer
(June–August) 2007,mean temperature and total precipitation
in northern Arizona were 178C and 13.8 cm (Western
Regional Climate Center 2008). Speciﬁc sites used for trials
were in areas managed primarily for forest and wildlife values
and were not in proximity to major roads or trails or to
recreational sites or other human structures.
Two handlers (authors Chambers and Mering) were trained
and certiﬁed to work with 2 conservation dogs at the
PackLeader Dog Training facility (Gig Harbor, WA).
Because we wanted dogs to respond to all bat species they
detected and not to speciﬁc bat species, dogs were trained to
locate bats using a mixture of guano from 5 species (pallid bat
[Antrozous pallidus], big brown bat [Eptesicus fuscus], Arizona
myotis [Myotis occultus], Yuma myotis [M. yumanensis], and
Mexican free-tailed bat [Tadarida brasiliensis]) collected at
natural roosts in Arizona and New Mexico, USA. Training
occurred at the facility in Washington; dogs and handlers
received identical training at the same time.
Following their arrival in northern Arizona, dogs had a
week to acclimate to elevation and climatic conditions before
performing trials from mid-July through mid-August 2007.
During this acclimation period, we used preliminary trials in
which we introduced dogs to guano at known roosts of
individual bat species and our guano mixture and became
conﬁdent that dogs responded to both the mixture of guano
and to individual species. We also introduced dogs to novel
species not included in the guano mixture (e.g., Allen’s
lappet-browed bat [Idionycteris phyllotis]). Dogs proved
equally adept at identifying individual species and novel
species. Dogs and their handlers (dog–handler pairs)
remained the same throughout preliminary trials and the
formal study. When a handler was not working with a dog,
she served as the orienteer for the working dog–handler pair.
We used 2 approaches to test the ability of dogs in ﬁnding
bat roosts: 1) experimental trials with simulated bat roosts
(i.e., bags of guano); and 2) ﬁeld tests to locate known bat
roosts (identiﬁed through radiotelemetry and exit counts)
that were currently or recently used (<1 month prior to ﬁeld
tests) by bats. For experimental trials, we tested effects of
roost location (height above ground of a guano source) and
amount of guano on ability of dogs to locate roosts. We
mixed and measured guano, in 5-g or 20-g quantities, into
organza fabric bags (7.6 10.2 cm). We compared detection
of simulated high versus low roosts using guano bags placed
at 6 m and 2 m above ground, respectively. We hung bags
from natural branch stubs adjacent to the bole, using a pole to
reach the 6-m placements. Our large concentration of guano
(20 g) simulated a large bat colony; the 5-g small
concentration simulated a small colony. We used empty
fabric bags as controls (“null bags”) to test whether dogs used
human scent associated with hanging the bags as an incorrect
indicator of guano.
We used 26 natural roosts for 6 Allen’s lappet-browed bat,
8 Arizona myotis, and 12 long-legged myotis (Myotis evotis)
that were occupied during the summer of 2007 and identiﬁed
through radiotelemetry during concurrent studies (Johnson
2008, Solvesky and Chambers 2009). We randomly assigned
a 1-ha plot around each roost so the position of the roost
relative to plot boundaries differed for each plot to prevent
For the approaches described above, we used 4 tests
(Table 1). The ﬁrst test (17–27 July) compared ability and
accuracy of a dog–handler pair in ﬁnding a 20-g guano
sample suspended 2 m above ground in a tree, a null bag (0 g
guano) as control, and known snag roosts. The second test
(30 July to 6 August) examined factors affecting the ability of
a dog–handler pair in ﬁnding a known snag roost for bats.
Because of the success of dogs in ﬁnding 20-g guano bags, we
decreased guano mass to 5 g for tests 3 and 4. This also better
simulated the small size of maternity colonies for roosts that
we used (xnumber of bats/roost was <20; Johnson 2008,
Solvesky and Chambers 2009). Thus our third test (7–13
August) compared ability of a dog–handler pair in ﬁnding a
5-g guano sample suspended 2 m above ground, a 5-g sample
suspended 6 m above ground, a null bag, and a known snag
roost used by bats. The fourth test (7–16 August) tested
accuracy of a dog–handler pair in ﬁnding 5-g guano samples
suspended at 2 heights (2 m or 6 m above ground).
For Tests 1 (n¼14) and 3 (n¼5), one dog–handler pair
searched in random order each of 3 1-ha plots containing
either a snag roost, guano bag, or null bag; and a trial
(sampling unit) was not complete until the dog–handler pair
searched all 3 plots. The orienteer randomly placed either a
guano or null bag in 2 of the 3 plots and the third plot
Chambers et al. Using Detection Dogs to Locate Bat Roosts 781
contained a snag roost whose position relative to the plot
boundary was unknown to the dog handler. We alternated
dog–handler pairs for each sequential trial. Average search
time per dog–handler pair per trial including rest breaks was
approximately 180 min.
Tests 2 (n¼36) and 4 (n¼16) involved only one 1-ha
plot/trial. We randomized which dog–handler pair started
the ﬁrst search each day and then alternated dog–handler
pairs. Average search time per dog–handler pair was 40 min.
During all tests, we recorded temperature, humidity, and
wind speed to test effects of weather on the ability of each
dog to accurately locate roosts or guano samples.
For all tests, we selected a random start point at one of the
1-ha plot corners, then searched each plot by walking 5 100-m
transects spaced 20 m apart. Dogs were unleashed but directed
by their handler along ﬁxed transect lines, and so were not
allowed to roam freely. The dog–handler pair focused on
searching with the handler closely observing the behavior of
the dog. The exposure of dogs to wind direction varied because
we used a random startlocation at one end of the 5 transectlines
and the direction of survey alternated between lines.
For tests 1, 3, and 4, dogs were correct if they indicated (sat
on their own, showing the handler that the dog had
concluded he found guano) 5 m from the known roost or
tree with the guano bag, and if they did not indicate a null
bag. Dogs were incorrect if they indicated >5 m from, or
failed to indicate, the known roost or tree with the guano
bag, or indicated on a null bag. For test 2, we compared poor
quality (imprecise and nondetections where the dog
indicated at >5 m or did not ﬁnd the object) with high
quality detections (dog pinpointed or indicated at 5 m from
the snag roost).
We collected information to compare hypothetical costs of
locating roosts with a dog or with radiotelemetry over a
1-month period. We based costs on previous studies we
conducted. Published studies using telemetry typically identify
30 roosts to describe habitat use (e.g., Bernardos et al. 2004,
Solvesky and Chambers 2009, Poissant et al. 2010), so for the
purpose of our cost comparison, we budgeted acquisition of
telemetry equipment (i.e., receiver, antenna; Communication
Specialists, Inc., CA) and 30 transmitters (Holohil Systems
Ltd., ON, Canada) using prices from companies that we
commonly used. Some expenses were recurring (e.g.,leasing a
dog, buying transmitters); others were one-time costs (e.g.,
handheld Global Positioning System [GPS] unit, receiver).
For the telemetry approach, we assumed the need for 3
technicians to capture and radiotag bats at night and 1
technician to locate day roosts. For the dog approach, our
budget accounted for a technician and an orienteer. We
assumed one of the technicians for each survey approachwould
conduct exit counts with equal costs per approach. We
included a required, 3-week course to train one technician in
dog handling. We also includedlease, care, and maintenance of
a dog. We assumed similar on-site travel for both approaches
but included round-trip travel to a facility for training for the
dog-handler technician and round-trip transport of the dog to
the study site.
To determine whether dogs were able to differentially detect
known roosts, 20-g guano samples, or null bags (test 1) and
accuracy of dogs in detecting 5-g guano samples at 2 heights
above ground (2 or 6 m; test 4), we used a comparison of
multiple proportions with all possible pairwise comparisons
(Goodman 1964, Fleiss et al. 2003). For test 2, we used
binary logistic regression to determine the best model(s) for
predicting correct identiﬁcation of a known bat roost by dogs
(SAS 9.3; SAS Institute, Inc. 2002–2012). We hypothesized
that the probability of a dog–handler pair correctly detecting
a roost was inﬂuenced by weather and characteristics of the
roost itself. We tested effects of roost (height above ground
[m], number of bats [determined from an exit count at the
roost] and weather (wind speed [km/hr], relative humidity
[%], temperature [8C]). We included dog–handler pair in
models as a ﬁxed effect but removed it if we detected no
difference between dog–handler pair. Because of small
sample sizes, we limited models to 3 variables and selected
P0.1 for the likelihood ratio chi-square statistic to test
signiﬁcance of models. We used percent correct classiﬁcation
(Percent Concordant) to compare accuracy of logistic
regression models. We assessed model ﬁt using a Hosmer
and Lemeshow goodness-of-ﬁt test (Hosmer and Lemeshow
2000). Sample size was small for test 3; therefore, we did not
statistically analyze data but instead presented descriptive
Because we found no differences between detection rates for
the 2 dogs (likelihood ratio chi-square ¼1.4, P¼0.2), we
combined trials across dogs. Dogs were more effective in
Table 1. Tests used to determine effectiveness of dog–handler pairs in locating known bat roosts (of variable ht above ground) and guano samples in
ponderosa pine forest, northern Arizona, USA, July–August 2007.
Test Objective N
No. of 1-ha
(min) Ht (m) 0g 5 g 20 g Roost
1 Ability to detect bat roosts and guano 14 3 180 2 X — X X
2 Factors affecting ability to find bat roost 36 1 40 Varied — — — X
3 Ability to detect roosts and guano at 2 ht 5 3 180 2, 6 X X — X
4 Accuracy in detecting guano at 2 ht 16 1 40 2, 6 — X — —
X indicates guano sample was used in the test, — indicates sample was not used.
782 Wildlife Society Bulletin 39(4)
detecting known bat roosts when dog–handler pairs alternated
searching a single 1-ha plot (test 2: 66% of roosts found within
5 m, 80% within 30m) than when they consecutively searched
3 1-ha plots (test 1: 29% of roosts found).
Test 1: Snag Roost, Guano Bag, or Null Bag
We completed 14 trials (7/dog). Dog–handler pairs correctly
avoided null bags 100% of the time, indicating that dogs were
not using human scent left during placement as a surrogate
for guano. Detection success rates were 79% for 20-g guano
bags and 29% for snag roosts (x
¼17.6, P<0.001; Table 2).
For 5 of the 10 incorrect detections for snag roosts, we noted
that dogs changed behavior or indicated a roost at a distance
>5 m. For the remaining 5 roosts we observed no change of
behavior for dogs.
Test 2: Factors Affecting Snag Roost Detection
Each dog–handler pair completed 17 or 19 trials (n¼36) for a
range of bat roost population sizes, heights, and weather
conditions (Table 3). High-quality detections (with the dog
pinpointing or within 5 m of a known roost) accounted for 64%
of trials. Imprecise detections included dogs identifying a roost
>5–20 m away (14%) or missing roosts entirely (22%). Roost
characteristics (number of bats that occupied the roost, height
of the roost) and roost characteristics in combination with
weather variables accurately predicted the ability of a dog to
locate a roost (Table 4). Wind speed, although statistically
signiﬁcant, was too inaccurate by itself in predicting a true
positive response (P¼0.05, % correct classiﬁcation ¼62.1).
The other weather variables (relative humidity, temperature)
did not contribute signiﬁcant explanatory power in single-
variable models (P>0.2, % correct classiﬁcation 64.8). All
weather variables were effective in increasing accuracy of
models when combinedwith roost height but onlywind speed
was effective with number of bats occupying a roost.
Combinations of temperature or relative humidity with
number of bats occupying a roost failed the Hosmer and
Lemeshow goodness-of-ﬁt test (P<0.02). All reported
models (Table 4) met the Hosmer and Lemeshow good-
ness-of-ﬁt test (P>0.05).
Models predicted strong, correct responses by dogs. The
model with greatest explanatory power was a 2-variable model
that included roost height and temperature. Dogs detected
roosts closer to the ground or with larger numbers of bats
(Fig. 1). Dogs also were more likely to detect roosts that were
close to the ground when temperatures were warmer(Table 4).
Tests 3 and 4: Amount of Guano and Height of Bag
Dogs were more accurate at detecting a smaller sample (5 g)
of guano when it was closer to ground level. They were
correct in ﬁnding guano samples placed 2 m high 60% of the
time but found guano samples placed 6 m high only 20% of
the time. They correctly detected a roost 60% of the time
(Fig. 2). In test 4, dogs were again more likely to ﬁnd guano
when it was closer to the ground. Dogs correctly identiﬁed
69% (11 of 16) guano samples at 2 m but only 40% of guano
samples (6 of 15) at 6 m above ground (P¼0.06).
On 35 occasions during our project, dogs gave an alert signal to
indicate the presence of guano at a site that was unknown to us
as a roost. Sites included 27 snags, 5 scats, 2 live trees, and 1 log
that could be false positives, previously used sites, or current
roosts. We suspect that 5 were misidentiﬁcations because dogs
alerted on scat of bobcat or coyote (Canis latrans) and we could
not visually conﬁrm the remains of a bat in the scat.
Cost Comparison—Dogs and Telemetry
Costs for the dog approach were 7% greater than use of
radiotelemetry (Table 5) for a 1-month hypothetical
sampling period. Personnel costs were greater for use of a
dog (83% of costs) than telemetry (44%); however, telemetry
required greater expenditures in supplies and equipment
(56%) than use of a dog (2%). Travel accounted for 15% of
costs associated with the dog approach. Both methods
included initial, one-time project costs (e.g., handheld GPS
unit, telemetry equipment, training in dog handling if the
same handler is used) but other costs would recur (e.g.,
Table 3. Means, standard error, and ranges for variables included in
models predicting the probability of a dog–handler pair finding a known
tree or snag roost for bats in ponderosa pine forests, northern Arizona,
USA, July–August 2007.
Variable nxSE Range
Roost ht (m) 17 11.0 1.7 1.5–30.0
Bat count (no. of bats exiting roost) 18 10.4 2.6 1–39
Temp (8C) 34 21.0 0.5 14.3–26.1
Wind speed (km/hr) 34 1.6 0.3 0–5.8
Relative humidity (%) 34 69.0 2.3 42–96
Table 2. Accuracy of the dog–handler pairs in locating (within 3, 1-ha plots) a 20-g guano sample suspended 2 m above ground in a tree, a null bag (no
guano), and a known snag roost in northern Arizona, USA, July 2007. Different letters indicate significant differences in detection among samples (x
P0.009), the same letter indicates no detectable difference among samples (x
Response of dogs
Percent correct Pairwise comparison
Guano 11 3 78.6 B
Null 14 0 100.0 B
Roost 4 10 28.6 A
Correct responses for Guano or Roost—The dog indicated on, or was 5m from, the tree or snag with guano bag or roost. Correct response for Null—The
dog ignored the bag with no guano.
Incorrect responses for Guano or Roost—Either the dog did not indicate, indicated on an incorrect location >5 m from tree or snag with the guano bag or
roost, or had a moderate to strong change of behavior but did not indicate a ﬁnd. Incorrect responses for Null—Dog indicated on the bag with no guano (this
Chambers et al. Using Detection Dogs to Locate Bat Roosts 783
personnel, expenses for the dog, transmitters, airfare).
Recurring costs for the dog approach were lower (59% of
total expenses) compared with telemetry (93%). Cost for
using a dog could be reduced if the dog handler also served as
the orienteer (e.g., from US$14,650 to US$12,250).
Although dogs have proven to be an effective tool for locating
live animals, carcasses, and scats of many species (Dahlgren
et al. 2012), our studyindicated that dogs were less effective at
locating bat roosts in snags, most likely because of the arboreal
location of the search object. When dogs are trained to locate
terrestrial species, the dog can provide an alert signal on the
targeted animal, carcass, or scat because the object is literally
under its nose (Wasser et al. 2004, Harrison 2006, DeMatteo
et al. 2009, Dahlgren et al. 2012). In contrast, a snag roost does
not provide a deﬁnitive terrestrial spot for the dog to indicate a
ﬁnd, unless there is guano below the roost. Werarely observed
guano below roosts, possibly because it was rapidly removed by
ants (Family Formicidae) and other insects. Height of snag
roosts increases the likelihood that guano scent will be carried
by the wind, which could confound the ability of the dog to
indicate the precise location of the scent. Dogs were most
successful at ﬁnding largeroosts and those closer to the ground.
Given that dogs showed a greater ability to detect guano
that was near the ground or in greater quantities, the species
of bat also may inﬂuence ability of dogs to detect roosts.
Larger bodied bats that roost closer to the ground (e.g.,
Allen’s lappet-browed bat, 12-g mass, with roosts typically
10 m above ground; Solvesky and Chambers 2009) or species
that aggregate in large colonies (e.g., Arizona myotis, Rabe
Table 4. Parameter estimates (b), standard errors (SE), and odds ratios (Exp(b)) for variables used in logistic regression models to predict probability of a
dog–handler pair finding a known tree or snag roost for bats in ponderosa pine forests, northern Arizona, USA, July–August 2007. Variables in models
included characteristics related to bats (bat count, ht of the roost) or weather (wind speed). The likelihood ratio tested significance of models, and the
c-statistic (area under the receiver operating characteristic curve) indicates a model’s ability to differentiate between outcome categories and the probability of
predicting a true positive response.
Bat count (no. exiting) 28 0.155 0.086 1.167 5.686 0.02 71.3
Roost ht (m) 24 0.153 0.087 0.858 3.834 0.05 77.8
Bat count, wind speed 26 7.168 0.03 83.5
Bat count 0.141 0.098 1.152
Wind speed (km/hr) 0.543 0.432 1.721
Roost ht, humidity 22 7.044 0.03 89.4
Roost ht 0.228 0.117 0.796
Relative humidity (%) 0.092 0.062 0.912
Roost ht, temp 22 6.623 0.04 92.9
Roost ht 0.241 0.123 0.786
Temp (8C) 0.377 0.257 1.458
Roost ht, wind speed 22 5.770 0.06 91.8
Roost ht 0.138 0.096 0.870
Wind speed (km/hr) 0.613 0.577 1.847
Sample size for characteristics related to bats varied because exit counts, ht of roost, or wind speed were unknown for some roosts.
Figure 1. Predicted probability of detection by a dog–handler pair for roost
height and number of bats occupying a roost in ponderosa pine forest,
northern Arizona, USA, July–August 2007.
Figure 2. Results of test 3 showing the percent correct (at or 5 m from
search object) or incorrect (no ﬁnd, dog uncertain, or >5 m from object)
searches for dog–handler pairs in a ponderosa pine forest, northern Arizona,
USA, August 2007. Dogs searched 3 consecutive 1-ha plots: 1) null bag with
0 g of guano (Null); 2) known snag roost (Roost); and 3) a 5-g guano bag
hung 2 m above ground (Guano2M) and a 5-g guano bag hung 6 m above
784 Wildlife Society Bulletin 39(4)
et al. 1998; C. Chambers, unpublished data) produce more
scent and thus should be easier to detect by dogs.
The live trees and log where dogs indicated the presence of
guano scent could represent misidentiﬁcations (i.e., false
positives), although male bats will roost in branch stubs
protruding from downed logs (Bernardos 2001) and tree-
roosting bats such as hoary and silver-haired bats (Lasiurus
cinereus and Lasionycteris noctivagans, respectively) that are
commonly captured in our study area use live trees. Most
snags identiﬁed by dogs appeared likely to be roost habitat
because they were large ponderosa pines (xdiameter ¼66
cm, xheight ¼16 m; e.g., Rabe et al. 1998, Bernardos et al.
2004, Solvesky and Chambers 2009). We conﬁrmed 8 as
current roosts because of the presence of guano or bats, but
we were unable to conﬁrm roost status for other snags (e.g.,
no visual or sound of bats). Bats switch roosts frequently
(e.g., every 7–14 days; Bernardos et al. 2004, Solvesky and
Chambers 2009), and may use roosts only during part of a
season or every few years. This can allow buildup of scent of
guano that is detectable to a dog although bats are not
present. The problem of unconﬁrmed roost sites is unlikely
to be resolved easily. However this could be an advantage
because dogs can locate roost sites that have been used in
the recent past and have high potential for reuse even though
the roost might be currently unoccupied.
Weather conditions affect scent detection (Shivik 2002,
Cablk et al. 2008, Reed et al. 2011); however, dogs have been
used successfullyin a variety of environments including tropical
moist forest (DeMatteo et al. 2009), temperate forest (Long
et al. 2007b), and desert (Cablk et al. 2008). In general, warmer
temperatures can release scent, concentrating it at the source
and thus making it stronger and easier for dogs to detect
(Shivik 2002, Cablk et al. 2008, Reed et al. 2011). Our work
was conducted in the American Southwest during monsoon
season, when relative humidity was high compared with dry
seasons. We had better success when temperatures were
warmer, humidity was lower, and light breezes prevailed.
Moister and windier conditions can broadly disperse scent at a
source too high or far away for dogs to detect or detect
accurately (Shivik 2002, Cablk et al. 2008, Reed et al. 2011);
therefore, the optimal conditions in our study likely distributed
scent lower in the canopy near the roost.
The cost of using dogs was not prohibitive, and as with any
method, had advantages and tradeoffs (Table 6). We found
that using a dog was similar in cost to radiotelemetry, could
allow land managers or researchers to concentrate a search in
a focal area, and was noninvasive and independent of the
need to capture and radiotag bats, which is an activity that
can be difﬁcult in some environments and particularly for
Table 5. Comparison of cost for domestic (U.S.) use of dogs and radiotelemetry to locate 30 bat roosts during a 1-month period, during trials conducted at 2
sites on the Coconino National Forest in northern Arizona, USA, during July and August of 2007.
Category (cost US$) Dogs Telemetry
Telemetry technician ($120/day, 20 days/month, 1 month) $2,400
Mist-netting technicians (3 @ $120/day, 10 days/month, 1 month) $3,600
Dog handler ($15/hr, 40 hr/week, 1 month) $2,400
Orienteer ($15/hr, 40 hr/week, 1 month)
Training in dog handling (3 weeks)
Dog lease ($2,000/month) $2,000
Dog transport and equipment ($250/month) $250
Dog food ($60/month) $60
Dog health certificate ($100/flight for 2 flights) $200
Insurance for dog lease (15% of lease) $300
Equipment and supplies
Transmitters (30 at $180 each, Holohil Systems Ltd.) $5,400
Mist nets (4 18-m nets at $139.75 each, 2 12-m nets @ $110.25 each)
Mist-netting equipment (poles, stakes, ropes, lights)
Garmin GPSMAP 62St Handheld GPS Navigator
R-1000 Telemetry Receiver
Yagi type directional antenna
Nondirectional magnetic mount antenna
Dog cargo airfare
Dog transport to/from airports ($25/trip) $50
Dog handler room and board during training ($250/week) $750
Dog handler airfare to or from Seattle
Total costs $14,650 $13,664
With handler only (no orienteer), cost for dog approach would be $12,250.
Assumed one-time cost (nonrecurring).
Assumed cost is $500/ﬂight for 2 trips to and from the study site.
Estimated cost for airplane ticket purchased 6 months in advance to and from central United States (Chicago) and Seattle, WA, for 2 weeks training at
PackLeader Dog Training facility.
Chambers et al. Using Detection Dogs to Locate Bat Roosts 785
Changing some cost assumptions could increase the cost of
telemetry over the use of dogs. We assumed a minimum
number of 30 transmitters to locate 30 roosts, which may be
underestimated depending on the study area and goals. We
assumed one-time costs for some equipment (e.g., receiver,
antenna); other equipment (e.g., mist nets) might need
replacement every 2–4 years. With dog surveys, the costs of
airfare, retraining for the dog handler (required when leasing
a dog after a hiatus in use), or hiring a professional handler
could all be much higher for certain survey locations. These
costs, however, could be avoided if the dog belongs to the
handler rather than a leasing service. Arnett (2006) provides
an example of a personally owned dog successfully trained to
Regardless of cost, land managers and researchers are
interested in the success rate of different survey methods
relative to sampling effort (e.g., Long et al. 2007a). A
comparison of success rates between radiotelemetry and dog
surveys in locating snag roosts is difﬁcult, however, because
of high variability in environmental conditions that affect
survey outcomes for each method. With radiotelemetry,
success rate can approach 100%, and telemetry can provide
the ability to double the number of known roosts if bats
change roosts while radiotagged (e.g., Bernardos et al. 2004,
Solvesky and Chambers 2009). In our trials with dogs,
success rate varied with height of roosts and to a lesser extent,
weather conditions, and we predict that sampling effort
could also vary widely if dogs frequently indicate roosts that
are not active but require additional survey time, such as an
evening exit count, to verify.
Whereas radiotelemetry is useful in all forest types, dogs are
likely to be most effective in short-statured woodlands such
as pinyon pine and perhaps certain hardwood forests,
although their usefulness in forests of moderate height (e.g.,
ponderosa pine and mixed conifer) should not be out ruled.
Taller roosts, however, increase the likelihood that dogs will
provide inaccurate alert signals or completely miss the roost.
Additionally, dogs are more likely to locate large bat colonies
than small ones.
Our data indicate that the costs of dogs and telemetry are
similar; therefore, the choice of method will depend on the
speciﬁc research question or management objective (Table 6).
Dogs can increase the sample of known bat roosts because
many species of bats use the same roosts repeatedly over
multiple years (e.g., Solvesky and Chambers 2009) and
guano can accumulate and make roost locations easier to
detect. Dogs can be trained to differentiate between species
(e.g., Harrison 2006, DeMatteo et al. 2009) and thus could
search for species that are uncommon, rare, or listed as
threatened or endangered, unless the species roosts too high
for scent detection. Prior to using dogs for any targeted
species, we recommend 2 weeks of preliminary trials and
training, to ensure that dogs can differentiate the scent of the
target species from other bats in the survey area.
Dogs could be useful for other arboreal and patchily
distributed species as well. Hagell (2010) experimented with
using a dog to locate an arboreal and locally rare primate (the
Central American spider monkey [Ateles geoffroyi]) and
found the dog method produced a sample yield of monkey
detections similar to that of human searchers during a
2-week pilot project. However, the dog identiﬁed 7 samples
not detected by the human team. Additional training time
and ﬁeld preparation beyond 2 weeks could have improved
performance of the dog. A minimum 2-week training period
is recommended when using dogs to acclimate them to a new
environment and to the local diet of the study organism.
Locating arboreal targets may require increased awareness of
scent movement, as might subterranean scent targets
(Reindl-Thompson et al. 2006).
Dogs can play a role in promoting awareness and
conservation of snag-roosting bats. At least 16 the 45
species of bats in Canada and the United States roost in
cavities or under bark (Barclay and Kurta 2007). If bat roosts
could be located in an area prior to forest management
activities (e.g., thinning or prescribed burning) or land use
changes (e.g., housing developments or ski area expansions),
managers could either protect existing roosts or provide
surrogate roosts as part of the mitigation strategy for the
planned activity (Mering and Chambers 2014). Although
dogs are not able to precisely locate all snag roosts, we found
they are generally within 30 m of the roost site, thereby
providing a targeted area for more concentrated searches or
When managers have the latitude to retain all snags in a
management area that meet roost characteristics, neither
dogs nor telemetry will be needed. Under those circum-
stances, managers should identify roost characteristics from
the literature and set standards for retaining trees and snags
Table 6. Ability of dogs and radiotelemetry methods to meet different aspects of the overall objective of locating snag roosts of bats in a targeted area during
trials conducted the study at 2 sites on the Coconino National Forest in northern Arizona, USA, during July and August of 2007.
Aspects of objective Dogs Telemetry
No. of bat species Can simultaneously search for roosts of multiple species or be
trained to search for roosts of one species
In general, can only locate roosts for the species that is
Type of roost Can focus entire effort on snags Unfocused; includes all roost structures
Spatial extent All roosts will be within a targeted area Roosts could be outside of the targeted area
Temporal extent Can locate previous, reusable roosts in addition to current
Can only locate currently used roosts
Degree of accuracy Might be 30m from the actual roost Can identify exact location of roost (assuming roosts are
Verification of roost Requires follow-up exit survey to determine if roost is still in
Requires follow-up exit survey to determine ht of roost and
786 Wildlife Society Bulletin 39(4)
that meet those characteristics. If only a subset of potential
roost sites can be retained in an area because of salvage
logging, hazard tree reduction, or other management
activities, dogs could be useful in identifying trees and
snags with the highest likelihood of current use. They could
also be used for postmanagement monitoring, to verify that
bats are indeed using retained structures as roost sites. In
comparison to radiotelemetry, dogs offer a less invasive
approach at a comparable price, which can be particularly
beneﬁcial when locating roosts of species that are imperiled
or of conservation concern.
We thank the U.S. Department of Agriculture Forest Service
and U.S. Department of Interior National Park Service for
funding. A. Chung-MacCoubrey provided impetus for the
project. We thank R. King (U.S. Forest Service) and Roy St.
Laurent (Northern Arizona University) for statistical
consulting help. Associate Editor Arnett and 3 anonymous
reviewers provided comments that greatly improved the
Aldridge, H. D., and R. M. Brigham. 1988. Load carrying and
maneuverability in an insectivorous bat: a test of the 5% “rule” of radio
telemetry. Journal of Mammalogy 69:379–382.
Amelon, S. K., D. C. Dalton, J. J. Millspaugh, and S. A. Wolf. 2009.
Radiotelemetry: techniques and analysis. Ecological and behavioral
methods for the study of bats. Pages 57–77 in T. A. Kunz and S.
Parsons, editors. John Hopkins University Press, Baltimore, Maryland,
Arnett, E. B. 2006. A preliminary evaluation on the use of dogs to recover
bat fatalities at wind energy facilities. Wildlife Society Bulletin
Barclay, R. M. R., and A. Kurta. 2007. Ecology and behavior of bats roosting
in tree cavities and under bark. Pages 17–59 in M. J. Lacki, J. P. Hayes, and
A. Kurta, editors. Bats in forests: conservation and management. Johns
Hopkins University Press, Baltimore, Maryland, USA.
Bernardos, D. A. 2001. Use of ponderosa pine-Gambel oak forests by bats in
northern Arizona. Thesis, Northern Arizona University, Flagstaff, USA.
Bernardos, D. A., C. L. Chambers, and M. J. Rabe. 2004. Selection of
Gambel oak roosts by southwestern myotis in ponderosa pine-dominated
forests, northern Arizona. Journal of Wildlife Management 68:595–601.
Cablk, M. E., J. C. Sagebiel, J. S. Heaton, and C. Valentin. 2008. Olfaction-
based detection distance: a quantitative analysis of how far away dogs
recognize tortoise odor and follow it to source. Sensors 8:2208–2222.
Chambers, C. L., M. J. Herder, K. Yasuda, D. G. Mikesic, S. M. Dewhurst,
W. M. Masters, and D. Vleck. 2011. Roosts and home ranges of spotted
bats (Euderma maculatum) in northern Arizona. Canadian Journal of
Corbett, R. J. M., C. L. Chambers, and M. Herder. 2008. Roosts and
activity areas of Nyctinomops macrotis in northern Arizona. Acta
Dahlgren, D. K., R. D. Elmore, D. A. Smith, A. Hurt, E. B. Arnett, and
J. W. Connelly. 2012. Use of dogs in wildlife research and management.
Pages 140–153 in N. Silvy, editor. Wildlife techniques manual. Volume 1.
Seventh edition. The Wildlife Society, Washington, D.C., USA.
DeMatteo, K. E., M. A. Rinas, M. M. Sede, B. Davenport, C. F.
Arg€uelles, K. Lovett, and P. G. Pareker. 2009. Detection dogs: an
effective technique for bush dog surveys. Journal of Wildlife Manage-
Fleiss, J. L., B. Levin, and M. C. Paik. 2003. Statistical methods for rates and
proportions. Third edition. John Wiley and Sons, New York, New York,
Goodman, L. A. 1964. Simultaneous conﬁdence intervals for contrasts
among multinomial populations. Annals of Mathematical Statistics
Gsell, A., J. Innes, P. de Monchy, and D. Brunton. 2010. The success of
using trained dogs to locate sparse rodents in pest-free sanctuaries.
Wildlife Research 37:39–46.
Harrison, R. 2006. A comparison of survey methods for detecting bobcats.
Wildlife Society Bulletin 34:548–552.
Hagell, S. E. 2010. Conserving forest connectivity for the Central American
spider monkey (Ateles geoffroyi) in southwestern Nicaragua. Dissertation,
Northern Arizona University, Flagstaff, USA.
Hosmer, D. W., and S. Lemeshow. 2000. Applied logistic regression.
Second edition. John Wiley and Sons, New York, New York, USA.
Johnson, S. A. 2008. Effects of ponderosa pine forest restoration on bat
habitat in northern Arizona. Thesis, Northern Arizona University,
Long, R. A., T. M. Donovan, P. Mackay, J. S. Buzas, and W. J. Zielinski.
2007a. Comparing detection dogs, cameras, and hair snares for surveying
carnivores. Journal of Wildlife Management 71:2018–2025.
Long, R. A., T. M. Donovan, P. Mackay, W. J. Zielinski, and J. S. Buzas.
2007b. Effectiveness of scat detection dogs for detecting forest carnivores.
Journal of Wildlife Management 71:2007–2017.
Mathews, F., M. Swindells, R. Goodhead, T. A. August, P. Hardman,
D. M. Linton, and D. J. Hosken. 2013. Effectiveness of search dogs
compared with human observers in locating bat carcasses at wind-turbine
sites: a blinded randomized trial. Wildlife Society Bulletin 37:34–40.
Mering, E. D., and C. L. Chambers. 2014. Thinking outside the box: a
review of artiﬁcial roosts for bats. Wildlife Society Bulletin 38:741–751.
Nussear, K. N., T. C. Esque, J. S. Heaton, M. E. Cablk, K. K. Drake, C.
Valentin, J. L. Yee, and P. A. Medica. 2008. Are wildlife detector dogs or
people better at ﬁnding tortoises? Journal of Herpetological Conservation
and Biology 3:103–115.
Paula, J. M., C. Leal, M. J. Silva, R. Mascarenhas, H. Costa, and M.
Mascarenhas. 2011. Dogs as a tool to improve bird-strike mortality
estimates at wind farms. Journal for Nature Conservation 19:202–208.
Poissant, J. A., H. G. Broders, and G. M. Quinn. 2010. Use of lichen as a
roosting substrate by Perimyotis subﬂavus, the tricolored bat, in Nova
Scotia. Ecoscience 17:372–378.
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.
Reed, S. E., A. L. Bidlack, A. Hurt, and W. M. Getz. 2011. Detection
distance and environmental factors in conservation detection dog surveys.
Journal of Wildlife Management 75:243–251.
Reindl-Thompson, S. A., J. A. Shivik, A. Whitelaw, A. Hurt, and K. F.
Higgins. 2006. Efﬁcacy of scent dogs in detecting black-footed ferrets at a
reintroduction site in South Dakota. Wildlife Society Bulletin
Rolland, R. M., P. K. Hamilton, S. D. Kraus, B. Davenport, R. M. Gillett,
and S. K. Wasser. 2006. Faecal sampling using detection dogs to study
reproduction and health in North Atlantic right whales (Eubalaena
glacialis). Journal of Cetacean Research and Management 8:121–125.
SAS Institute, Inc. 2002–2012. SAS 9.3: help and documentation. SAS
Institute, Inc., Cary, North Carolina, USA.
Shivik, J. A. 2002. Odor-absorptive clothing, environmental factors, and
search-dog ability. Wildlife Society Bulletin 30:721–727.
Solvesky, B. G., and C. L. Chambers. 2009. Roosts of Allen’s lappet-
browed bat in northern Arizona. Journal of Wildlife Management
Vice, D. S., and R. M. Engeman. 2000. Brown tree snake discoveries during
detector dog inspections following Supertyphoon Paka. Micronesica
Wasser, S. K., B. Davenport, E. R. Ramage, K. E. Hunt, M. Parker, C.
Clarke, and G. Stenhouse. 2004. Scat detection dogs in wildlife research
and management: application to grizzly and black bears in the
Yellowhead Ecosystem, Alberta, Canada. Canadian Journal of Zoology
Weller, T. D., P. M. Cryan, and T. J. O’Shea. 2009. Broadening the focus of
bat conservation and research in the USA for the 21
Species Research 8:129–145.
Western Regional Climate Center. 2008. Historical climate information.
Temperature and precipitation for Flagstaff, Arizona. http://www.wrcc.
dri.edu/index.html. Accessed 9 Sep 2008.
Associate Editor: Arnett.
Chambers et al. Using Detection Dogs to Locate Bat Roosts 787