Variation in Search-Phase Calls of Bats

Abstract

Although echolocation calls of most bats exhibit species-specific characteristics, intraspecific variation can obscure differences among species and make reliable acoustic identification difficult. We examined levels of intraspecific variation in search-phase calls of 7 species of vespertilionid bats from several locations in the eastern and central United States. Echolocation calls were recorded from light-tagged bats using the Anabat II detector and associated software. Analook software was used to calculate values for 5 parameters of calls: duration, maximum frequency, minimum frequency, frequency of the body, and slope of the body. Analysis of our results indicates that most intraspecific variability in calls was attributable to differences among individuals and within individual call sequences. Observed levels of geographic variation, although significant in all species examined, were comparatively small and showed no trends among areas. We include a preliminary description of variability in echolocation calls of Nyticeius humeralis and Myotis leibii.

Anabat, Analook, bats, echolocation, geographic variation, intraspecific variation, Myotis, Vespertilionidae

Insectivorous bats use echolocation to obtain information about their surroundings, including location and identity of potential prey items (Kalko and Schnitzler 1989; Obrist 1995). Typical sequences of echolocation calls consist of search-phase calls used to detect a prey item, approach-phase calls used in pursuit, and terminal-phase calls (feeding buzzes) used just before capture (Griffin et al. 1960; Kalko and Schnitzler 1989). Search-phase calls are ideal for studies of acoustic identification of species for several reasons. Search-phase calls are emitted more often by foraging bats than are other types of calls and consequently are encountered frequently in the field (Fenton and Bell 1981; Fenton and Morris 1976). Search-phase calls also exhibit consistency in structure throughout the call sequence (Fenton and Bell 1981; O'Farrell et al. 1999) and usually have species-specific characteristics (Ahlén 1981; Fenton and Bell 1981; O'Farrell et al. 1999).

Although search-phase calls of a species may be distinctive, intraspecific variation can obscure differences among species and make identification problematic (Barclay 1999; Betts 1998; Brigham et al. 1989; Thomas et al. 1987). Echolocation calls can differ among individuals because of size, age, and sex (Buchler 1980; Heller and Helversen 1989; Jones et al. 1992), or because individuals are foraging in different habitat or microhabitat types (Jacobs 1999; Jensen and Miller 1999; Rydell 1990, 1993). Structure of individual pulses within a call sequence can change as bats maneuver within their environment or as they encounter obstacles or conspecifics (Kalko and Schnitzler 1993; Obrist 1995; Rydell 1990, 1993). Variation in search-phase calls among populations from different geographic regions also has been demonstrated for some species (Barclay et al. 1999; Parsons 1997; Thomas et al. 1987). These factors combined with effects of atmospheric attenuation, Doppler shift, directionality of emitted signal, and directionality of the detector can lead to high levels of intraspecific variation (Obrist 1995; Parsons et al. 1997; Pye 1993), which translate into an increased potential for overlap of calls among species. Thus, reliable characterization of interspecific differences in structure of echolocation calls may require an extensive understanding of levels of intraspecific variation (Barclay 1999; Betts 1998; Brigham et al. 1989; Thomas et al. 1987).

Most studies have focused on intraspecific variation among and within individuals because of differences in habitat or foraging strategy. Comparatively few studies have examined variation among populations from different geographic locations (Barclay et al. 1999; Jones and Kokurewicz 1994; Thomas et al. 1987). These studies often examined small numbers of individuals of 1–2 species, depended on echolocation calls recorded in buildings or enclosures, or failed to determine degree of variation. Thomas et al. (1987) compiled data from 20 species of bats from different geographic regions in North America and demonstrated that geographic variation was possible. However, because that study relied on data from several researchers using different detectors and methods of analysis and because parameters of echolocation calls often were measured from only a few individuals, the amount of intraspecific variation that exists among populations is still unclear.

Our purpose was to describe intraspecific variation in search-phase echolocation calls of 7 species of vespertilionids found in the eastern and central United States. In particular, we characterized variation in species over a broad geographic area by using consistent methods and by examining a relatively large sample of individuals. We attempted to quantify various levels of intraspecific variation including variation among and within individuals and among populations. Our analysis focused on geographic variation because the extent to which species vary across their geographic range remains unclear and because several authors have cautioned that geographic variation may preclude identification of echolocation calls of species from different regions (Barclay 1999; Brigham et al. 1989; de Oliveira 1998; Thomas et al. 1987).

Materials and Methods

We captured 9 species of bats (Eptesicus fuscus, Lasiurus borealis, Myotis grisescens, M. leibii, M. lucifugus, M. septentrionalis, M. sodalis, Nycticeius humeralis, and Pipistrellus subflavus) using harp traps and mist nets in 6 geographic areas of the eastern United States (Fig. 1): area 1, southwestern Missouri (Christian, Dallas, Greene, and Taney counties) and southeastern Kansas (Crawford County); area 2, eastern Missouri (Iron and Washington counties); area 3, central Indiana (Greene County) and western Kentucky (Livingston County); area 4, central Kentucky (Pulaski and Edmonson counties) and central Tennessee (Putnam and Fentress counties); area 5, eastern Tennessee (Blount County) and western North Carolina (Graham County); and area 6, central Michigan (Manistee County). Small chemiluminescent tags (Chemical Light, Inc., Vernon Hills, Illinois) were attached to the dorsal side of each bat (Buchler 1976; Hovorka et al. 1996), and bats were transported to a predetermined release site. Bats usually were released in large open areas (100–300 m in diameter) within 0.5 km of the capture location. However, in some cases, bats were released in open flyways along streams (most bats from areas 2 and 6). To decrease likelihood of recording echolocation calls of unknown, free-flying individuals, bats were released at sites with minimal activity of bats as determined by previous sampling with Anabat II. Recording stations were placed about 50 m from the release point. That distance was sufficient to allow bats to acclimate to their surroundings and begin to produce search-phase calls before reaching a recording station. That was somewhat analogous to recording calls away from a roost because emerging bats do not emit typical search-phase calls until they are away from the roost site (Jones and van Parijs 1993).

Fig. 1.

Location of the 6 geographic areas in the east-central United States from which echolocation calls of bats were compared

Recording stations consisted of an Anabat II bat detector, a Zero Crossing Analysis Interface Module (ZCAIM), and a laptop computer running the Anabat software (Titley Electronics, Ballina, New South Wales, Australia). The Anabat II detector is a broadband (20–200 kHz), divide-by detector, that divides frequency of the incoming signal by a preset division ratio (set at 16 in this study). That brought the signal into audible range and allowed for analysis of calls. The signal then passed through the ZCAIM where frequency information for the harmonic with the most energy (usually the fundamental) was determined by zero-crossing analysis. With this type of analysis, multiple harmonics were excluded from the recorded echolocation call. The ZCAIM was connected to a laptop computer running Anabat software (versions 6.2d or 5.7i), which constructed a frequency–time graph of the detected echolocation call and allowed calls to be saved to the hard drive for later analysis.

To supplement samples of species that were infrequently captured or recorded at some locations, 2 additional methods of obtaining known calls were used. Echolocation calls of E. fuscus and M. grisescens were recorded from free-flying bats near the entrance of known single-species roosts. In addition, most of the known search-phase calls from L. borealis were recorded from bats flying at dusk.

Echolocation calls were analyzed using the software program Analook (version 4.7j, Titley Electronics). Typical sequences of echolocation calls contained predominately search-phase pulses but also contained approach-phase and terminal-phase pulses (feeding buzzes), echoes, extraneous noise, and fragmentary pulses. We quantitatively removed non–search-phase portions of the call sequence using the filter option in Analook. The filter automatically edited an echolocation call sequence and removed some of the subjectivity that may have existed when calls were edited manually (Barclay 1999; Britzke and Murray 2000). Additional cleaning sometimes was necessary to remove remaining fragmentary calls, and that was done using the Mark Off Points options in the edit menu of Analook.

When the editing process was completed, values of parameters for each pulse within a call sequence were saved for further statistical analysis. Edited call sequences with <5 pulses were excluded from analysis. We used 5 individual parameters calculated by the Analook software to examine intraspecific variation in search-phase calls of the 9 species of bats. We examined 3 frequency components, maximum frequency, minimum frequency, and frequency of the body; 1 time component, total duration of the pulse; and 1 slope component, slope of the body. For clarification, body of the call was defined as the flattest part of the call (lowest frequency change over time) and was equivalent to the narrow-band component of the call (Caddle and Lumsden 1997).

Statistical tests were performed on 7 species of bats (E. fuscus, L. borealis, M. grisescens, M. lucifugus, M. septentrionalis, M. sodalis, and P. subflavus) using the Minitab-version 12.23 software package (Minitab Inc., State College, Pennsylvania). A nested multivariate analysis of variance was performed to examine intraspecific variation in search-phase echolocation calls among individuals from 6 geographic regions in the eastern and central United States (Fig. 1). A nested analysis of variance was then used to examine levels of variation among geographic regions, among individuals, and within-individual call sequences for each individual call parameter. A single sequence of echolocation calls was considered to be an individual bat, and thus, within-individual variation referred to differences among single calls within a call sequence. The nested analysis of variance generated variance components that revealed the percentage of total variation explained by differences among geographic areas and individuals. The error-variance component described amount of variation within individuals. Because of the unbalanced design, a general-linear model was used to conduct significance tests (α = 0.05) for interindividual and geographic variation for each call parameter. Tukey multiple-comparison tests (α = 0.05) were then performed to determine which geographic regions differed from one another (Sokal and Rohlf 1981). A linear discriminant-function analysis using individual means for all 5 call parameters also was used to determine if it was possible to assign echolocation calls of each species to correct geographic area.

Echolocation calls of N. humeralis and M. leibii were recorded from a single location and represent only 1 or a few individuals. Thus, we reported only means and coefficients of variation (CVs) for those 2 species.

Results

Table 1 lists total number of individuals (call sequences) and pulses from each species and geographic area included in our analysis; no species was recorded from every geographic area. Means of parameters and CVs for each species (Table 2) were similar to values reported in previous studies (Betts 1998; Brigham et al. 1989; Fenton 1994; Fenton and Bell 1981; MacDonald et al. 1994; Obrist 1995; Thomas et al. 1987). Means and CVs for N. humeralis and M. leibii are included in Table 3. Coefficients of variation showed that slope of the body was the most variable parameter, probably because of relatively high inherent variability and inconsistencies in its measurement by the Analook software (C. Corben, pers. comm.), followed by duration and maximum frequency. Minimum frequency and frequency of the body exhibited a comparatively small amount of variation (Table 2).

Table 1.

Number of individual call sequences recorded from each species for each geographic area. Numbers in parentheses represent total number of pulses recorded from all individuals in a particular area (Fig. 1)

Table 2.

Means (CVs in parentheses) of 5 parameters of calls for each bat species from each geographic area; units are duration (ms), maximum frequency (kHz), minimum frequency (kHz), frequency of the body (kHz), and slope of the body (octaves/s). Areas with the same letter are not significantly different from each other (Tukey multiple-comparison test). For explanation of geographic areas, see text and Fig. 1

Table 3.

Means (CVs in parentheses) for 5 parameters of calls recorded from 4 Nycticeius humeralis and 1 Myotis leibii

Nested multivariate analysis of variance showed significant variation among individuals and among geographic locations. The general-linear model showed similar significant variation for all parameters for each species with the only exception being frequency of the body in M. grisescens (Table 4). According to the nested analysis of variance and corresponding variance components, the most common source of variability occurred within individual call sequences, with differences among individuals also accounting for a relatively high percentage of total variation (Table 4).

Table 4.

Results of nested analysis of variance for each species of vespertilionid bat. Variance components show the relative amount of variation (%) among geographic areas, among individuals, and within individuals for each call parameter. Asterisk (*) indicates significant variation (P < 0.001); † denotes negative variance components that were rounded to zero and considered significant by Minitab Statistics Package, version 12.23. NS = not significant

Geographic variation was significant in nearly all cases and accounted for a high percentage of total variation in 2 species, E. fuscus and L. borealis (Table 4). However, variance components showed that variation due to geographic area was relatively low in the 5 other species examined (Table 4). Tukey multiple-comparison tests revealed that significant variation among geographic locations was common for all species and parameters (except frequency of the body in M. grisescens). However, no trend in variation among geographic locations was apparent (Table 2). No individual site was consistently different from all others, variation between sites did not increase as distance between sites increased, and no clinal variation was apparent.

Despite significant geographic variation, linear discriminate-function analysis was unable to reliably assign individual sequences of echolocation calls to geographic area. Overall rates of accuracy were ≤64.6% in 6 of 7 species examined (Table 5). Only calls of L. borealis were classified somewhat accurately (82.8%), probably because of the relatively high levels of geographic variability in this species.

Table 5.

Combined resubstitution results of discriminant-function analysis for 7 bat species using 5 parameters of calls (duration, maximum frequency, minimum frequency, frequency of the body, and slope of the body) testing ability to assign calls of a species to the correct geographic area (Fig. 1). Data in columns represent number of individuals assigned to the proper geographic area out of the total number of bats recorded in that area; the final column indicates total accuracy for a species with overall rate of accuracy in parentheses

Discussion

Our results indicate high levels of variation in search-phase echolocation calls for all 7 species of bats examined. Coefficients of variation for duration, maximum frequency, and minimum frequency are similar to those reported in several studies (e.g., Betts 1998; Brigham et al. 1989; Fenton 1994; Obrist 1995,). In addition, ranges of values for species that have been well studied, such as L. borealis, E. fuscus, and M. lucifugus, fall within limits of previously published values (Brigham et al. 1989; Fenton 1994; Fenton and Bell 1981; Thomas et al. 1987). This indicates that observed levels of variation in recorded calls are representative of actual variability that exists in emitted echolocation calls. Nevertheless, some amount of the observed variation is inevitably a result of inconsistencies in recording situation. In most instances, bats were released singly in areas with low levels of activity and in similar habitat types (large open areas), minimizing variation due to presence of conspecifics (Jones et al. 1994; Miller and Degn 1981; Obrist 1995) or differences in habitat (Kalko 1995; Miller and Degn 1981; Schumm et al. 1991). However, this was not always the case. Our analysis included calls of free-flying individuals in less-controlled conditions near known roost sites, above ponds at dusk, and over open-stream flyways (areas 2 and 6). These factors could have contributed to overall variation in calls.

A portion of the observed variation could be attributable to the Doppler effect or atmospheric attenuation. Doppler effect can cause a shift in recorded frequency as a bat is flying toward or away from the detector at different speeds. Thus, Doppler effect may have elevated levels of variation among individuals or within an individual sequence of calls (as a bat flies toward, over, and away from the detector). Jones and van Parijs (1993) attributed consistent intraindividual changes in frequency of <2 kHz in Pipistrellus pipistrellus to Doppler shift. In our study, CVs for minimum frequency and maximum frequency, the parameters most likely to be affected by Doppler shift, are similar to those reported by Obrist (1995) under controlled conditions, indicating that effects of Doppler shift were minimal. Atmospheric attenuation, absorption of sound energy by the air, becomes more extreme as the frequency of sound increases (Griffin 1971; Lawrence and Simmons 1982). Typically, atmospheric attenuation is used to explain relatively high levels of variation in maximum frequency, because the higher frequency portion of the echolocation pulse is expected to be attenuated more rapidly (Fenton and Bell 1981; Parsons et al. 1997; Pye 1993; Thomas et al. 1987). Coefficients of variation for values of maximum frequency from all species range from 4% to 15% and are similar to previously reported values for maximum frequency (Bellwood and Fullard 1984; Betts 1998; Obrist 1995). These values are consistently higher than those for the other 2 frequency components, minimum frequency and frequency of the body (Table 3). Thus, atmospheric attenuation may be responsible for some of the variation seen in maximum frequency.

Intraspecific variation among and within individuals has several possible biological explanations. Many studies have shown that size, age, and sex could contribute to the observed variability (Heller and Helversen 1989; Jones and Ransome 1993; Moss et al. 1997). Kalko and Schnitzler (1993), using stroboscopic photography in conjunction with acoustic analysis, showed that individuals can alter their echolocation calls as they change their flight path or encounter clutter. Although most individuals in our study were released in open areas, bats did approach surrounding vegetation and may have altered calls as a result. The high levels of variation in duration and slope of the body probably are due in part to these factors. Variability also may be indicative of individual-specific echolocation calls, which would allow bats to differentiate between their own echoes and those of a conspecific (Fenton 1994; Jones et al. 1992; Masters et al. 1995; Obrist 1995).

The absolute effect that geographic variation had on echolocation calls of these 7 species is difficult to determine from our results. The fact that significant overall variation, as well as significant variation for all species and nearly all parameters, was observed is a strong indication that geographic variation exists. Tukey multiple-comparison tests also showed that significant differences between geographic areas were common (Table 2). However, analysis of our results also indicated that other sources of variation had a much greater effect on echolocation calls than did geographic area. Variance components showed that geographic location often accounted for only a small percentage of total variation and, in almost all cases, variability among and within individuals was a more significant contributor to overall variation (Table 4). The discriminant-function analysis demonstrated that sufficient variability did not exist among areas for 6 of 7 species to accurately classify calls to the proper geographic location (Table 5). These factors, along with the relatively small amount of variation observed, make the biological significance of geographic variation for acoustical analysis questionable for the 7 species examined.

Parsons (1997) reported geographic variation in echolocation calls of 2 species of bats in New Zealand (Mystacina tuberculata and Chalinolobus tuberculatus). Variation in calls of M. tuberculata corresponded fairly well with subspecific designations, whereas variation in calls of C. tuberculatus followed clinal morphologic differences from northern to southern New Zealand. Jones and Kokurewicz (1994) documented variation in calls of Myotis daubentonii from different locations in Europe and proposed that variation may be related to previously documented morphologic variation for the species (Bogdanowicz 1990). These studies indicate that variation in structure of echolocation calls may be related to gross morphologic variability. Other possible causes of geographic variation include genetic or learned differences among populations, varying foraging habitats or prey items, or morphologic variation in vocal structures used to emit echolocation pulses (Brigham et al. 1989; de Oliveira 1998; Jones and van Parijs 1993; Obrist 1995). To date, only Barclay et al. (1999), looking at variation in calls of 2 subspecies of Lasiurus cinereus, have examined any of these proposed hypotheses. We showed that significant geographic variation existed in several North American species, but we were unable to document any probable cause for this variation.

Our results document considerable intraspecific variability in echolocation calls and have important implications for future research. Any studies concerning acoustic identification of bats should rely upon an extensive understanding of intraspecific variation for species being studied. This is certainly not a novel idea because several authors have made similar recommendations (Barclay 1999; Brigham et al. 1989; de Oliveira 1998; Thomas et al. 1987). However, our findings contradict these recommendations somewhat in regards to geographic variation. For most species we studied that geographic variation apparently is slight and intraspecific variation can be described adequately with a large sample of individuals (30–100) from a small geographic area. Therefore, accurate identification of species from a broader geographic area using known calls obtained from only a few locations is possible. However, having some understanding of specific structure of echolocation calls from the study locality is still advisable. Finally, high levels of within-individual variability mandate that future studies of geographic variation in structure of echolocation calls should use a large sample of individuals from the populations being studied. If only a few individuals are sampled, any apparent disparity among populations from different geographic areas may simply be an artifact of high levels of individual variation. Undoubtedly, caution must be exercised when conducting research regarding acoustic identification or geographic variation until intraspecific variation can be described from a sufficiently large sample of individuals for the species in question.

Acknowledgments

We thank the United States Fish and Wildlife Service, United States Forest Service, and Missouri Department of Conservation for providing funding for this study. The Department of Biology at Southwest Missouri State University and the Center for Utilization, Protection, and Management of Water Resources at Tennessee Technological University provided valuable support to the project. We also thank the United States Forest Service, Missouri Department of Conservation, and Mammoth Cave National Park for access to lands under their jurisdictions. Bat Conservation International and M. J. O'Farrell played a critical role in recording echolocation calls in Indiana. We express our gratitude to C. Corben for technical advice on the use of the Analook software, J. S. Heywood for aiding in statistical analysis, and R. Holmes and W. Goodman of Southwest Missouri State University Support Services for providing quality field equipment for the study. Finally, we thank all those individuals that assisted us in the field, especially B. Hadley, D. Bossi, L. Solberg, J. MacGregor, M. Harvey, S. K. Amelon, and R. Currie, without whose help the project could not have been completed.

Author notes

* Correspondent:

Associate Editor was Troy L. Best.