OUP user menu

Tree-Roosting Ecology of Reproductive Female Eastern Pipistrelles, Pipistrellus subflavus, in Indiana

Jacques Pierre Veilleux, John O. Whitaker Jr., Sherry L. Veilleux
DOI: http://dx.doi.org/10.1644/BEM-021 1068-1075 First published online: 29 August 2003


We studied roosting ecology of reproductive (pregnant or lactating) adult female eastern pipistrelles (Pipistrellus subflavus) in Indiana. Nineteen bats were radiotagged and 37 roost trees identified. Pipistrelles roosted exclusively in foliage, typically in clusters of dead leaves (65%) and less often in live foliage (30%) or squirrel nests (5%). Oaks (Quereus species) were preferred as roost trees. Roost trees and height of foliage roosts were both located well below the forest canopy. Bats remained at roost trees for 6 days on average before moving to new roosts and traveled approximately 19–139 m between roost trees. This is the 1st thorough analysis of roosting habits of this common species under natural conditions.

Key words
  • bats
  • Chiroptera
  • eastern pipistrelle
  • habitat selection
  • Indiana
  • Pipistrellus subflavus
  • reproduction
  • roost
  • site fidelity

Many species of temperate North American bats have experienced population declines, mainly due to loss or disturbance of suitable roosting habitats in summer or winter. Gestation and rearing of young occur within prematernity and maternity roosts during late spring and early summer. To develop conservation strategies and management plans, it is important to have a thorough understanding of a species' natural history, particularly the characteristics of habitats (Fenton 1997) and roosts used by breeding females. However, little information on roosting ecology of breeding females is available for most North American bat species, especially in natural versus human-related habitats.

The eastern pipistrelle (Pipistrellus subflavus) is a common bat of the eastern United States (Barbour and Davis 1969). Most studies of the species' summer-roosting habits have been of colonies in man-made structures (Allen 1921; Cope et al. 1961; Jones and Pagels 1968; Jones and Suttkus 1973; Whitaker 1998; Winchell and Kunz 1996).

Pipistrelles seem to use man-made structures less often than syntopic species. For example, pipistrelles accounted for only 12 of 401 (2.9%) bat colonies in buildings in Indiana (Cope et al. 1991), although the species is relatively common there (11.5% of 6,445 bats captured in a state-wide mist netting survey from 1990 through 2000; J. O. Whitaker, Jr., in litt.). In addition, pipistrelle colonies in buildings are small (mean size =15 adult females—Whitaker 1998) compared with colonies of syntopic species in buildings (e.g., means for big brown bat, Eptesicus fuscus = 151 and for little brown bat, Myotis lucifugus = 564— Mumford and Whitaker 1982).

The relative scarcity of the eastern pipistrelle in buildings in Indiana suggests that most colonies are in forests or at least in trees. However, there is little information on the species' roosting habits in woodland habitats. Because eastern pipistrelles occasionally form colonies in buildings, we postulated that they would form their roosts within tree hollows as indicated by Davis and Mumford (1962) and Menzel (1996). Davis and Mumford (1962) noted that a museum specimen was from a hollow tree stump. Further, Menzel (1996) collected (on separate days) a lactating and a juvenile female in a pitfall trap placed in front of a basal cavity of a sweet gum tree (Liquid-ambar styraciflud) in Georgia; the cavity may have been used by a maternity colony. Other records indicate that pipistrelles will roost in leaves (Findley 1954; Veilleux et al. 2001) or Spanish moss (Tillandsia us-neoidesDavis and Mumford 1962; Menzel et al. 1999). Finally, Carter et al. (1999) and Kurta et al. (1999) radiotracked single adult females to several roost trees in late summer. In both studies, it appeared as though the bats were emerging from high in the upper canopy foliage. In summary, pipistrelles may roost in foliage or hollows.

The objective of this study was to describe tree-roosting ecology of breeding (pregnant or lactating) adult female pipistrelles. Specific objectives were to determine locations of and substrates used at tree roosts, characteristics of roost trees and the surrounding habitat, and roost-site fidelity.

Materials and Methods

This study was conducted from late April through July, 1998–2000, mainly at Prairie Creek, in southwestern Vigo County, Indiana (UTM 43.48.000N, 4.54.000E). Prairie Creek is a small tributary of the Wabash River and is located in the glaciated section of the Southwestern Lowlands Physiographic Region (Homoya et al. 1985). The eastern portion of the study site (Fig. 1) contains approximately 180 ha of noncontiguous upland woods, situated along the creek edge, and nearby small, isolated woodland fragments. The woods are dominated by sugar maple (Acer saccharum), sycamore (Platanus occidentalis), hackberry (Celtis occidentalis), tulip tree (Liriodendron tulipifera), and American elm (Ulmus americand). From the upland woodland, the creek flows E through approximately 1.5 km of agricultural fields. This stretch of creek is bordered by about 15 ha of riparian woodland, 20–40 m wide on both north and south sides. This riparian vegetation is composed mainly of silver maple (Acer sacchari-num), hackberry, osage orange (Madura pomi-fera), and hawthorn (Crataegus). The creek then turns S and enters the western portion of the study site, composed of a 650-ha hardwood bottomland forest, surrounded by agricultural fields of corn and soybeans. The main tree species within the bottomland forest are silver maple, green ash (Fraxinus pennsylvanica), cottonwood (Populus deltoides), and American elm.

Fig. 1

Prairie Creek study site in southern Vigo County, Indiana, showing upland, riparian, and bottomland woodland habitat types.

Due to small numbers of trees and to facilitate analysis of tree-species preference, tree species were combined into 3 categories: Quer eus (Q. alba, Q. macrocarpa, and Q. rubra),Acer (Acer negundo, A. saccharum, and A. saccharinum), and other species (Aesculus glabra, C. occidentalis, Fraxinus americana, Juglans nigra, L. tulipifera, Nyssa sylvatica, P. deltoides, and U. americana).

Most pipistrelles were captured by mist netting, using 2 mist nets, one above the other, for a total surface area of 45 m2 (9 m horizontal and 5 m vertical). Mist nets were placed across Prairie Creek at 5 primary sites and several additional sites. One bat was mist netted along Little Raccoon Creek at the Newport Chemical Depot in Vermillion County, Indiana (approximately 60 km N of Prairie Creek), and another was mist netted at Prairie Creek Park along a small creek (not associated with Prairie Creek) approximately 7 km E of the main study site. Mist nets were typically attended by observers from dusk until midnight. Nets were taken down early because of poor weather on several occasions. Additionally, 3 pipistrelles were captured by hand while researchers examined prospective roosts during the day. Mist netting was also used in conjunction with radiotelemetry data, to assess whether pipistrelles were more common in different habitat types within the study site (i.e., bottomland woods, narrow riparian corridors, or upland woods). A total of 22 net nights was obtained for upland habitats, 17 for riparian, and 15 for bottomland habitat areas.

Sex, age, body mass, and reproductive status were recorded for each bat. Age class (adult or juvenile) was determined by degree of ossification of epiphyseal plates in the finger bones (Anthony 1988). Adult females were classed as pregnant by degree of distention of the abdomen, and lactation was determined by whether milk could be expressed after gentle pressure was applied to the nipple (Racey 1988). Bats were considered to be nonbreeding if the abdomen was not distended (as is typical of bats early in the year, i.e., late April to early May) or if no milk could be expressed from nipples that obviously had been suckled (i.e., in the case of postlactating females). A numbered plastic or aluminum wing band (Lambournes Limited, West Midlands, United Kingdom) was fitted to each captured bat (left wing of females and right wing of males—Barclay and Bell 1988). A white plastic band was also fitted to the right wing of radiotagged females.

To locate roost trees of reproductive adult females, bats were fitted with small (0.45 g) ra-diotransmitters (Model LB-2, Holohil Inc., Carp, Canada; or Model LTM, Titley Electronics, Bal-lina, Australia). A small amount of fur was trimmed from between the scapulae using scissors, then the transmitter was glued into place using a nontoxic surgical adhesive (Skin-Bond, Smith & Nephew Inc., Largo, Florida). After gluing, bats were held for 3–5 min to allow the glue to dry. Bats were released at the point of capture. Radiotagged bats were tracked to their day roosts on the following day and on each day thereafter until the transmitter battery failed or the transmitter fell from the bat. In 1998 and 1999, only 1 bat was radiotagged at a time; in 2000, up to 4 bats were radiotagged concurrently

Bats were tracked using a radioreceiver (Model TRX2000S, Wildlife Materials Inc., Carbon-dale, Illinois) and 3-element y agi antenna. After a roost tree was located, up to 4 h were spent searching for the specific roost location using 8 by 40 binoculars and a 15–60× zoom spotting scope. Trees were considered to be verified as roosts only if the bats were seen in the roost during the day or when emerging at night. Nightly emergence counts were made at each roost to obtain data on daily variations in colony size and to allow observations of roost-changing activity. Emergence counts were conducted typically 15–20 min before emergence until 10 min after the last bat emerged. A video camera (Model AG-456UP, Panasonic., Secaucus, New Jersey) was used to tape record emergences at an additional roost, which allowed for additional census data on multiple roosts on the same night. On several occasions, personnel gathered census data on additional roosts on the same night. If bats were tracked to a roost tree on the 1st day of monitoring and on subsequent days never reused that roost, the tree was not included in later analyses, to guard against potential bias.

Several roost-tree and roost-site characteristics were recorded for each verified roost tree. Most were measured within a 0.1-ha (17.8 m radius) circular plot centered on the roost tree. All tree heights were measured using a clinometer (model PM5/360PC., Suunto, Carlsbad, California) or laser range finder (Impulse 200 Laser, Laser Technology, Inc., Englewood, Colorado). Horizontal distance to the nearest tree as tall as or taller than the roost tree was measured with a 30-m tape or laser range finder. Tree diameter at breast height was measured using a diameter tape measure (model 283D/5M, Forestry Suppliers, Inc., Jackson, Mississippi). Nine decay stages of roost trees were recognized (1, live-tree stage, through 9, dead fallen stump), following Thomas et al. (in litt.). Percentage canopy cover was defined as the amount of area above the roost proper (where the bats were specifically roosting) that prevented light from reaching the roost due to overhead limbs and foliage. Percentage canopy cover was estimated visually. Four estimates were taken, 1 in each of 4 quadrants. Quadrants were areas located above a 90° sector, which was then set at a 45° pitch, forming an imaginary cone above the roost. The 4 measures were averaged and the mean taken as the best estimate.

Percentage clutter above and below the roost was also estimated (Betts 1998). Percentage clutter above the roost was defined as the amount of woody material and foliage that broke an imaginary horizontal plane, set within an imaginary 3-m-long, 1-m-radius half cylinder, located immediately above the roost. Percentage clutter below was measured using the same method, except that the half cylinder was placed below the roost. A measure of general canopy height within the roost plot was obtained by averaging the heights of 3 randomly selected canopy trees.

Latitude and longitude readings were obtained at each roost tree using a Garmin 12XL global positioning system (GPS) handheld receiver (Garmin International Inc., Olathe, Kansas). Additional GPS points were obtained at the nearest permanent water source, the nearest forest edge relative to the roost tree, and the bat's capture site. GPS points were downloaded onto digital (TopoQuad software) United States Geological Survey quadrangle maps (Delorme Publishing Company, Inc., Yarmouth, Maine). All long-distance measurements were obtained using this software (i.e., distance between subsequent roosts, distance to permanent water and forest edge).

Circular statistics were used to determine if bats roosted at random with regard to roost aspect (Zar 1984). Chi-square analyses were conducted to determine selection of roost-tree species and habitat preference.

Four variables associated with fidelity were examined for each individual during the period it was monitored, including number of roosts used, number of changes of roost sites, maximum number of consecutive days spent at a single roost, and mean number of days spent at a roost. Data on roost-site fidelity for analyses were obtained from pipistrelles monitored for >7 days (range, 7–17 days).

Pairwise associations between roost-site fidelity, roost characteristics, and climatic variables were tested with Spearman's correlation coefficient. Mean daily high and low temperatures and rainfall were calculated for periods when bats were monitored. Results are presented as mean ± SD.


Nineteen breeding female pipistrelles (14 pregnant; 5 lactating) were radiotracked to 66 roosts during the study, and 37 roost trees were verified. All verified roosts were in foliage, including those with young bats present. Most foliage roosts were in clusters of dead leaves (n = 26), but some were in live foliage (n = 11). Dead-leaf clusters occurred as 4 different types: broken branches still attached to the original tree (n = 20); broken branches completely separated from original tree but settled lower in the canopy of the original tree or a nearby tree (n = 4); dead leaves formed by a natural cause (i.e., no obvious structural damage; n = 1); and an abandoned gray squirrel (Sciurus caro-linensis) nest (n = 1).

Shapes of dead-leaf roosts were similar regardless of size. Roosts took the shape of an umbrella, resulting in a protective roof of dead foliage and a hollow core into which bats retreated. Live-leaf roosts were dense in structure, but on 3 occasions the bats were found clasping a branch with only a few leaves serving as shelter.

Tree species and habitat areas.—Roosts occurred in 14 different tree species and were mainly in the Quercus and Acer classes (Table 1). More roosts occurred in Quercus and fewer in Acer and other species than expected (ϰ2 = 30.75, d.f. = 2, P < 0.001).

View this table:
Table 1

Use of tree species as roosts by pregnant and lactating eastern pipistrelles (Pip-istrellus subflavus) and proportions of tree species sampled within 0.1-ha roost plots.

Tree speciesRoosts, no. (%)Percent occurrence in plots
Quer eus alba2 (5.4)0.01
Q. rubra7 (18.9)3.1
Q. macrocarpa1 (2.7)0.02
Acer saccharum4 (10.8)18.1
A. saccharinum3 (8.1)10.8
A. negundo8 (21.6)8.8
Aesculus glabra2 (5.4)2.5
Nyssa sylvatica1 (2.7)2.8
Ulmus americana1 (2.7)6.5
Fraxinus americana1 (2.7)2.1
Celtis occidentalis2 (5.4)5.6
Juglans nigra2 (5.4)2.2
Liriodendron tulipifera1 (2.7)1.9
Populus deltoides2 (5.4)2.7

Thirty-six of the 37 roosts were in live trees. A single dead Q. rubra was used, which had approximately 30 small, dead-leaf clusters on its branches from the previous year. The tree appeared to have been struck by lightning.

Female pipistrelles were captured with equal frequency in each habitat type (x2 = 3.17, d.f. = 2, P > 0.05) but roosted more often than expected in upland and riparian habitat areas and less often in bottomland (ϰ2 = 56.86, d.f. = 2, P < 0.0001). Fourteen bats (73.7%) roosted in the upland area, 3 in the riparian area, and 2 in the bottomland area.

Measured characteristics of roost trees and roost plots.—Thirteen roost-tree and roost-site characteristics were measured. Roost-tree height (tree height, hereafter) averaged 20.8 m ± 7.1 SD (range, 4.7–37.6 m). Height from ground to the leaf cluster (roost height, hereafter) was 15.7 ± 6.8 m (range, 2.5–36.1 m). Diameter at breast height was 33.2 cm ± 18.8 SD (range, 4.4-83.3 cm).

Roost height relative to tree height was 74.7% ± 17.6 SD (range, 29–99%; i.e., the cluster of leaves averaged three-quarters of tree height above the ground). Tree height relative to general canopy height averaged 74.8 ± 26.4% (range, 16–120%), and roost height relative to canopy height averaged 56.5 ± 24.4% (range, 7–113%).

Percentage canopy cover averaged 41.4 ± 29.5% (range, 0–95%). Roosts (live and dead-leaf clusters) had little vegetation obstructing the areas immediately above and even less below the leaf cluster used by bats. Percentage clutter above the roost averaged 35.1 ± 25.7% and below the roost averaged 8.3 ± 13.3%. Mean roost aspect was 96.5°. Pipistrelles did not select leaf roosts with a southern exposure (U = 0.0003; P > 0.05), but roosts were randomly distributed along the circumferences of roost trees (z = 0.02; P > 0.05).

Distance to and height of the nearest tree as tall as or taller than the roost tree averaged 6 ± 5 m and 24 ± 8 m, respectively. Distance to water source and forest edge averaged 117 ± 154 m and 52 ± 47 m, respectively.

Site fidelity.—Estimates of roost-site fidelity were obtained for 18 bats. Bats were monitored for 9.1 days ±1.7 SD. Bats used 2.8 ± 1.7 roost trees and changed roost trees 2.3 ±1.9 times, on average. Bats remained at single roost trees for 6.0 ± 2.7 consecutive days on average, although 4 individuals that were in late pregnancy or lactating remained at a roost for up to 17 days. Overall, bats remained at a roost for 3.9 ± 2.5 days. Eight of 18 individuals (44.4%) returned to previously used roost trees after initially changing to a new roost.

Colony size was similar before and after roosts were changed, suggesting that colonies changed roosts as groups, although some variation was observed. For example, in 1999, a pregnant bat was monitored from 21 to 30 May, and colony size remained at 3 individuals after 4 of 5 roost changes. On the 2nd occasion, colony size for a pregnant pipistrelle monitored from 17 to 29 June remained at 3 individuals after 3 of 4 roost changes. Finally, colony size for a pregnant pipistrelle monitored from 14 to 22 May 2000 remained at 2 individuals after 3 roost changes.

Relationships of roost characteristics and climatic variables to roost-site fidelity.—No significant correlations between site fidelity and roost characteristics were detected. However, roost-site fidelity was significantly correlated with weather variables: maximal daily temperature was significantly correlated with number of roosts (r = −0.46, P = 0.046), number of roost changes (r = − 0.58, P = 0.010), and maximal number of consecutive days spent at a roost tree (r = 0.49, P = 0.033); minimal daily temperature was significantly correlated with number of roosts (r =−0.57, P = 0.010), number of roost changes (r = −0.67, P = 0.002), and maximal number of consecutive days spent at a roost (r = 0.62, P = 0.004).

Maximal number of consecutive days spent at a roost did not differ between dead-and live-leaf roosts (3.9 ±3.7 days in the former and 2.2 ± 2.4 days in the latter; Mann-Whitney test, U = 101.5; P = 0.283) or among the 3 roost-tree species categories (Quercus = 3.2 ± 2.5 days; Acer = 3.4 ± 2.8 days; other species = 3.3 ± 2.5 days; F = 0.018, d.f. = 2, 33, P = 0.983).

Twenty-five distances between successive roost trees averaged 60 ± 33 m (range, 19–139 m). Two pregnant and 4 lactating individuals used only single roost trees. Roost area for females that used ≥3 roost trees was 0.23 ha ± 0.15 SD (range, 0.03–0.42 ha; n = 8). Mean distance from the site of capture to roost area was 0.72 km ± 0.87 SD (range, 0.05–2.61 km; n = 15). The latter was not significantly correlated with time of capture (r = 0.121, P = 0.667), suggesting that bats did not travel greater distances as the night progressed.


Tree-roost characteristics.—All roosts used were within foliage, not tree hollows, as we had predicted. This makes the eastern pipistrelle unique among north-temperate bat species. Other north-temperate foliage-roosting species (Lasiurus) roost singly (except adult females with dependent young), and colonial species roost in more stable structures (typically available over multiple years) such as buildings, tree hollows, under exfoliating bark, or caves (Kunz 1982). Roosting in foliage may have evolved to reduce competition for roosting in tree hollows or under exfoliating bark because at least 8 sympatric bat species use tree hollows or bark as roost sites. Other competition for these sites (especially tree hollows) comes from other mammal and bird species.

Quercus species were important roost trees for pipistrelles in this study. Carter et al. (1999) lists 4 of 6 roosts used by a single adult female pipistrelle in Quercus (Q. mi-chauxii and Q. laurifolia), and a single juvenile pipistrelle radiotagged by Kurta et al. (1999) roosted in 2 trees, Q. alba and Q. rubra. Quercus typically has terminal branches with many leaflets clustering near the last several centimeters of the branch. When these branches break and hang downward, a tight cluster of dead leaves is formed, resulting in an umbrella-like shelter. The umbrella shape of roosts used by pipistrelles is similar to that observed by Constantine (1966) for temperate Lasiurus. It may provide shelter from wind and rain and may moderate roost microclimate. Bats within such clusters cannot be seen from above or below, hence are hidden from visual predators. Leaves of those Quercus species used as roosts also have broad leaf blades, which likely aid in concealment and provide protection from wind and rain (McClure 1942). Finally, Quercus leaf roosts are relatively stable over time compared with other roost-tree species.

Acer species were used as roosts less than expected but nevertheless were important as roost sites. The number of leaflets near the end of the terminal branches of A. saccharum and A. saccharinum is fewer than in Quercus, but petiole length is longer, so leaves originating far from the end of a branch may be included in the roost cluster. A. negundo was used relatively often, but pipistrelles exhibited low fidelity to this species. Most roosts in A. negundo were in live foliage, possibly because its leaves are compound and so offer little protection. The presence of dense clusters of samaras may offer some camouflage because it was difficult to distinguish them from bats.

Other tree species were used infrequently. P. deltoides and L. tulipifera have broad but widely spaced leaves; C. occidentalis, N. sylvatica, and U. americana have small, widely spaced leaves and short petioles (1 roost in U. americana was formed by a grape vine, Vitts). The compound leaflets of U. americana, J. nigra, A. glabra, and F. americana often formed rather dense leaf clusters but may provide little protection against wind.

Pipistrelles may have preferred upland habitat because of the relatively greater availability of preferred roost tree species there, especially Quercus (9.4% of upland trees versus 0.8% in riparian and 0.2% in bottomland habitats). It was expected that breeding females would roost at a high level, which would provide warm temperatures through exposure to sunlight (Betts 1998; Ormsbee and McComb 1998; Vonhof and Barclay 1996; Vonhof and Barclay 1997). However, we found that roost height relative to canopy height and tree height relative to canopy height were low. The reasons for this are unclear but may relate to exposure of the colony to reduced climatic extremes at low levels (Mager and Nelson 2001; McComb and Noble 1981; Smith and Smith 2000; Walsberg 1986; Wolf and Walsberg 1996). Avoidance of the forest and field edge may be due to similar effects (Davies-Colley et al. 2000; Gehlhausen et al. 2000).

Insolation is related inversely to the amount of clutter above a roost. Less clutter may also decrease the risk of predation, partly by providing an unobstructed flight path for escape (Campbell et al. 1996; Vonhof and Barclay 1996).

Roost fidelity.Pipistrellus subflavus exhibits greater roost-site fidelity than other temperate foliage-roosting species. For example, temperate Lasiurus borealis and L. seminolus spend about 1.2 and 1.7 days at a roost tree, respectively (Menzel et al. 1998), whereas pipistrelles remained for about 3.9 days (and some individuals in late pregnancy or lactation remained for up to 17 days).

At high ambient temperatures, female pipistrelles switched roost trees less often, used fewer roosts, and remained at roosts for longer periods. Interestingly, roost-site fidelity of reproductive females was not correlated with roost-tree or roost-site attributes. It would be valuable to assess effects of wind velocity.

Distances traveled by pipistrelles are similar to those of big brown bats. Brigham (1991) found that the latter species traveled 1.8 ± 0.1 km (maximum 4.4 km) from roost areas to foraging grounds; in our study, maximal distance traveled was 4.3 km.

Future work should document whether geographic variation exists in terms of roost use by eastern pipistrelles. For example, research should examine whether P. subflavus uses foliage as a roost substrate in other geographic areas throughout its range or shows flexibility in roost choice (i.e., roosts in hollows or under exfoliating bark). Additional support for roost-tree species preference (Quercus) and level of roost fidelity in other geographic areas would be worthy of investigation as well.


We thank the landowners living in and around the Prairie Creek study site, especially H. Clark, J. Strain, B. Evans, and J.-A. Evans, A. Krochmal, J. Duchamp, and C. Ritzi. We also thank 2 anonymous reviewers and T. H. Kunz for reviewing an early draft. We thank the Life Sciences Department of Indiana State University for financial assistance and use of field vehicles and equipment. Additional funding was provided by grants from Bat Conservation International, the American Museum of Natural History (Theodore Roosevelt Fund), the Indiana Academy of Science, and the Indiana State University Graduate School Research Fund to J. P. Veilleux.


  • Associate Editor was Edward H. Miller.

Literature Cited

View Abstract