Variation in the frequency of the echolocation calls of ...



Variation in the frequency of the echolocation calls of Hipposideros ruber in the Gulf of Guinea: an exploration of the adaptive meaning of the constant frequency value in rhinolophoid CF bats

A. GUILL E’ N, * t J. JUSTE B . t & C . I B A’ N EZ t

*Department of Biology, University of Missouri-St. Louis, 8001 Natural Bridge Road, St. Louis, MO 63121—4499, USA

tEstacio n Biolo gica de Don ana, CSIC, Apartado 1056, 41080 Sevilla, Spain

Departamento de Biouu mica y Biolog a Molecular I , Universidad Complutense de Madrid, aacultad de eterinaria, 28040 Madrid, Spain

Keywords:

adaptation; Chiroptera;

constant frequency; echolocation; geographical variation; Hipposideros ruber; humidity;

speciation.

Abstract

This study describes variation patterns in the constant frequency (CF) segment of echolocation calls of the bat Hipposideros ruber within and among populations across the region of the Gulf of Guinea. Correlations of variation in CF with variation in body size, body condition, environmental humidity and presence of ecologically similar species are studied in an attempt to identify the forces driving the evolution of CF. We found that bats may adapt the frequency to humidity, and that CF may evolve under interspecific interactions, either of ecological or of social nature. The results support an adaptive value for the high values of CF, and challenge the Allotonic Frequency Hypothesis’. We found correlation between frequency and a body condition index, which may trigger social selection processes in this species sexually dimorphic in CF. Combined social and environmental selection on CF could trigger diversification of bats along ecotones separating habitats with contrasting air humidity.

Introduction

Horseshoe (Rhinolophidae) and roundleaf (Hipposider- idae) bats use an echolocation system that relies on calls containing constant frequency segments (hereafter CF; Griffin, 1958). Constant frequency segments of these calls have much higher pitch than calls used by other bats, a characteristic associated with a shift from the first to the second harmonic as the information carrier. The bats construct complex sensorial images of their prey after micromodulations produced in the echo of the originally CF segment by the differential movements of the body parts of the fluttering insects they hunt (Kober

& Schnitzler, 1990; von der Emde & Schnitzler, 1990). Neurological, behavioural and ecological studies have recently increased our understanding of this narrow

Correspondence: Dr A. Guille& n, Department of Biology, University of Missouri-St. Louis, Saint Louis, MO 63121, USA. Tel.: +1 314 516 6207; fax: +1 314 516 6233;

e-mail: aguillen@admiral.umsl.edu

analysis sonar system, although the reason for using such high frequencies remains unclear (Fenton & Fullard,

1979; Fullard, 1987; Schnitzler, 1987; Vater, 1987; Heller

& Helversen, 1989; Neuweiler, 1989, 1990a; Ru” bsamen & Scha” fer, 1990a,b; Jones, 1995, 1996, 1997).

Echolocation requires high-frequency sounds that are highly directional and limit perception to unobstructed objects located at relatively short distance. Signal char- acteristics will evolve primarily in response to factors causing distortion of the signal in the direct pathway from the emitter to the target and back (attenuation, absorption), and factors affecting echo formation (char- acteristics and position of prey with respect to the solid background). Other factors important to the evolution of social signals, such as pattern loss by scattering, reflection and refraction of sound by objects in the transmission medium (Bradbury & Vehrencamp, 1998), should have little significance in evolution of echolocation.

Because echolocation is a short-range detection sys- tem, it is expected that bats that hunt for insects flying in open air will use, in calls or parts of the calls whose main

function is prey detection, the frequencies that give them the largest detection distances for the preferred prey. While the absolute value of the detection distance changes with air temperature and humidity, maximum echo strength in the aerial detection of a spherical target is always achieved when using a sound with wavelength n times the diameter of the target ([wavelength n X diameterj; Pye, 1983; Hartley, 1989). Then, an inverse relationship of frequency vs. ideal prey diameter is expected; and because predator and prey body sizes are directly related (Peters, 1983), an inverse relationship between frequency and bat body size is also expected. The main function of the ending quasi-CF segment of calls produced by aerial hawking insectivorous bats is prey detection (Simmons & Grinnell, 1988; Neuweiler,

1989, 1990). Frequency of this segment shows a negative trend towards body size with an elevation that suggests an adaptive match between frequency and prey size in those bats (Pye, 1983; Barclay & Brigham, 1991; Jones,

1996, 1997). Horseshoe and roundleaf bats also show an inverse relationship between CF and body size, although the intercepts are much higher (Heller & Helversen,

1989; Barclay & Brigham, 1991; Jones, 1996). It is clear from the studies cited above that these bats neither use the CF segment for mere detection nor hunt in open spaces (i.e. Neuweiler, 1989). Higher frequencies may be necessary when the goal instead is to encode vibrational characteristics of prey. The inverse trend could then arise when the resolving power of smaller moving structures increases with the frequency of the call carrying the information.

An alternative explanation, framed as the Allotonic Frequency Hypothesis’, states that high frequencies used by horseshoe and roundleaf bats are a way to circumvent the auditory defences of moths, which may make up the bulk of the bats’ diet (Fenton & Fullard, 1979; Fullard,

1987; Jones, 1992). Since frequency response of the

auditory organs of moths does not seem to correlate with moth size (Fenton & Fullard, 1979; Surlykke, 1988), the allotonic frequency hypothesis denies a functional mean- ing of the relationship between CF and body size. A developmental constraint (i.e. the coupling of devel- opment of structures for sound production or reception with the development of the skull) might then explain the trend (longer vocal chords and larger cavities produce and resonate sounds of lower frequencies).

A simultaneous demonstration of an adaptive value of lower CF and the absence of a tight constraint between frequency and body size would favour the functional meaning of the trend vs. the allotonic frequency expla- nation. Selective pressures favouring lower frequencies would lead, through adaptive evolution, to a clustering of the CF of all rhinolophid and hipposiderid bats just above the upper frequency threshold of the hearing system of tympanate moths. The existence of the developmental constraint may be investigated by comparing patterns of covariation between CF and body size among sex,

populations and species living under different social or ecological conditions.

Insights into the potential adaptive meaning of CF may be obtained from the evolutionary response of the character to changes in the assemblage of immediate ecological competitors. If CF value were related to prey size, its mean and/or variance would likely respond to changing interspecific competitive interactions for resource partitioning (Van Valen, 1965; Arthur, 1982; Endler, 1986; Jones, 1995, 1997), as would any other ecomorphological attribute (Grant, 1994).

High-frequency sounds are attenuated in the atmo- sphere at higher rate than deeper ones, and the rate at which absorption of sound energy increases is directly related to both air humidity and frequency (Lawrence & Simmons, 1982; Hartley, 1989). Perceptive ranges of rhinolophids and hipposiderids might be rather short, potentially limiting and bats experience great losses in range with increasing humidity (Hartley, 1989). When frequencies higher than needed for mere detection are used for other perceptive purpose, humidity may deter- mine a levelling point in a trade-off between selection towards higher frequencies to obtain more resolution in prey classification, and towards lower frequencies to achieve longer range. Correlation of a trait (frequency) with an environmental variable (air humidity at the hunting grounds) would also suggest the existence of natural selection on the trait (Endler, 1986).

Bats with CF calls, including rhinolophids and hippo- siderids, use echolocation calls, or signals structurally similar to the echolocation calls, in communication (Fenton, 1985). Given the formidable capacity of these bats for discriminating frequencies around the emitted CF (Vater, 1987), social information might be encoded in minute differences or modulations in frequency. This situation may exert strong selection on individuals to maintain their CF frequencies close to the population average in order to interact socially. Heller & von Helversen (1989) have suggested the existence of acoustic communication channels’ in these bats. Fre- quency might also represent a species-specific recogni- tion signal (Butlin, 1995). Information on the health of a bat might also be encoded in the frequency of its sounds (Huffman & Henson, 1993), which may allow social selection processes. Social selection may influence with- in-and between-population variation in CF, potentially confounding the effects of ecological factors, and should be considered in the study of the selective universe operating on the evolution of characteristics of echo- location calls in bats.

In this paper we study intraspecific patterns of varia- tion in the CF of Hipposideros ruber (Noack 1893) in the islands of the Gulf of Guinea (Central Africa) and the immediate mainland in an attempt to understand factors underlying variation in CF. These populations live syntopically with different numbers of congeneric species and offer the opportunity to test for the existence of

(1) a constraint of body size on CF, (2) shifts in means or changes in the variance or sexual dimorphism of CF indicating an adaptive variation of the trait associated with changes in the composition of the ecological assemblage, (3) adaptation of CF to local humidity and (4) covariation between CF and a body condition index, since a social role of the CF value could be rooted in the information about phenotypic or genotypic quality pro- vided by CF as a signal. Possible social and ecological causes underlying the patterns found and their evolu- tionary implications are discussed.

Materials and methods

The species and the geographical settings

The colonial bat Hipposideros ruber is widely distributed and abundant in the Central African rainforest belt and surrounding forest—savanna mosaics. On the mainland, this species lives syntopically with at least 10 other species of rhinolophid and hipposiderid bats. Hipposideros ruber is also present on the larger islands of the Gulf of Guinea (Hayman & Hill, 1971). The large land-bridge island of Bioko holds an important subset (six species) of the taxa present in the immediate mainland; whereas, only one congeneric species coexists with H. ruber on the oceanic island of Sa o Tome& , and none on the smaller oceanic island of Pri&ncipe (Juste B. & Iba& n ez, 1994). Although the whole region is included in the rainforest biome, local populations of H. ruber experience rather different environmental conditions. Conspicuous con- trasts occur between the southern slopes and the driest northern slopes of the central volcanic massifs of the islands with dramatic differences in sunshine, humidity and rainfall due to foehn’s effect (Fig. 1). These climate differences are impressively reflected in the vegetation, which changes from hyperhumid rainforest in the southern areas to deciduous forest in the northern areas, with baobab savannahs on some of the islands (Juste B.

& Fa, 1994). It may be inferred that the impressive differences in rainfall and vegetation structure are reflected in differences in average air humidity in the hunting grounds of bats living in different areas.

Data collection and sound analysis

Bats were netted in different expeditions from October

1992 to February 1994. A total of 437 bats from 16 colonies were sampled in R&io Muni (Equatorial Guinea, Western central Africa) and the islands of Bioko, Pri&ncipe and Sa o Tome& (Fig. 1). We intended to sample colonies across the environmental range that the different popu- lations experience, but due to logistic difficulties this plan was realized fully only on Sa o Tome& (Table 1, Fig. 1). Only fully grown specimens, with complete epiphysial fusion of finger joints (Anthony, 1988), were recorded for the study. Forearm length (FA), measured to the

nearest 0.1 mm with dial calipers, was used as a simple measure of body size. Body mass was measured to the nearest 0.5 g with Pesola’ spring scales for individuals from colonies 3, 5, 7 and 12.

Recordings of calls with resting CF frequency were obtained from hand-held bats restrained motionless

15 cm in front of the microphone of a Lars Pettersson

Elektronik D960 ultrasound detector. Bats with narrow analysis echolocation systems broadcast in this situation true echolocation calls, structurally indistinguishable from the calls produced in hunting flight. The resting frequency’ is the individually characteristic and stable CF emitted by the motionless bat, which coincides with the response frequency of the cochlear acoustic fovea’ (a greatly expanded segment of the basilar membrane specialized to resolve frequencies in a narrow interval around the CF of the call; Ru” bsamen & Scha” fer, 1990a). In contrast, flying bats broadcast calls with variable frequency compensated for the Doppler effect caused by movement, so the echo returns with a carrier frequency around the resting frequency’, allowing the narrow frequency analysis (Schnitzler, 1970). Signals were slowed-down 10 times with the detector built-in A/D— D/A converter and recorded onto metal-XR Sony tapes with a Sony WM-D6C cassette recorder. Recordings were analysed on a Kay DSP 5500 Sonagraph with the sampling frequency set to 40 kHz. The bats broadcast typical hipposiderid echolocation calls, composed of an initial CF segment 6—9 ms long followed by a steep downward frequency modulated sweep (Fig. 2). The frequency of the CF component in the second harmonic (the functional one) was measured from an average power spectrum built with 512-point fast Fourier trans- forms and taken over the complete CF segment (Fig. 2). The resolution attainable with this process was 400 Hz. For each individual, 10 calls were measured and the mean value was used in the analysis. Within-individual variation was very small, the SD averaging 0.3 kHz. For controlling shifts in frequency due to equipment failure or power instability, an 880-Hz reference sound from a musical tuner was recorded at intervals interspersed with data recordings. No significant departures from the reference frequency were detected in the analysis.

Hypotheses and statistical analyses

Variable names are typed in uppercase to avoid confusion with their general biological meaning. As a factor, POPULATION has four levels and it refers to the four different populations of the species sampled (three islands and the mainland). COLONY refers to each of the 16 colonies sampled (Fig. 1).

Because the distribution of both CF and FA did not depart significantly from normality, but data were unbalanced among cells, general linear models (GLM) were used for the analyses with SAS statistical package (Littell et al., 1991). Since some cells missed data, type IV

Fig. 1 Study area with the location of sampled localities. Altitude and precipitation iso-lines are according to Tera& n (1962) and Jones et al. (1991).

sums of squares were used to build the tests (Shaw & Mitchell-Olds, 1993). Full models were built initially, but they were simplified subsequently by removing nonsig- nificant terms. Parameter estimates produced by the option SOLUTION of GLM procedure of SAS were inspected to determine how factor levels contributed to the patterns detected.

Changes in means of CF related to the sexual and geographical structure of the populations (potentially related to changes in ecological assemblage) were studied with a linear model that included SEX and POPULATION as main fixed effects and COLONY as a random effect nested within POPULATION (nested effect typed as COLONY[POPj).

Since longer vocal chords and larger cavities produce and resonate sounds of lower frequencies, changes in body size could cause changes in CF if this trait lacked adaptive value or if it were severely constrained during ontogeny by body size. Variation in FA, a trait considered under strong selection and tight developmental control because of its relation with body size (Williams, 1992) and its fundamental role for wing performance (Gummer

& Brigham, 1995), was used as reference for assessing the importance of variation in CF. A GLM with the same structure as the one presented above was used for studying variation in FA. The relation between general body size and CF was assessed within colonies by studying the correlation between the residuals from the

Females Males

Pop C N mCF mFA N mCF mFA MAR

Table 1 Sample size (N) and mean ± SD of resting frequency (mCF, in kHz), and fore- arm length (mFA, in mm), per colony and

sex. Last column shows mean annual rainfall

s.

[pic]

Fig. 2 Sonogram (left) and power spectrum (right) of a typical echolocation call of Hipposideros ruber. Maximum energy is on the second harmonic, which is used as the information carrier.

CF and FA GLM models described above, both for all data and within sex.

The effects of POPULATION, COLONY and SEX on the variances of CF and FA were studied with a multilevel— multifactorial adaptation of Levene’s test (Van Valen,

1978). The absolute difference between each measure-

ment and the median of the corresponding cell (POPU- LATION X COLONY X SEX) was divided by the cell mean. The new variables (unsigned relative deviations) were used as response variables for two linear models with the same structure as those used for studying the population structure of original variables CF and FA. The

amount of variation in CF was compared with that in FA by a Wilcoxon paired-sample test on the corresponding unsigned relative deviations.

Since parallelism in development could yield correla- tion patterns between traits that are actually evolution- arily independent, it is relevant to check whether correlations detected within populations also hold across populations under different selective regimes. We simul- taneously studied the effects of FA and environmental humidity across populations with an A N C O V A model in which the means of CF and FA per colony (Table 1) were the response variable and covariate, respectively (mCF and mFA), and POPULATION was the main fixed effect. In the analyses described above we found significant correlation between CF and FA, but different slopes for each sex; therefore, SEX was also included as a main fixed effect in the model and a separate-slopes by SEX effect was specified for mFA (mFA[SEXj). To test for the effect of environmental humidity we included local mean annual rainfall (hereafter MAR) as a second covariate in the model. MAR was obtained (Table 1) from isoyet lines published in Tera& n (1962) and Jones et al. (1991). We consider this a rough index of the target variable: humidity at the hunting grounds of bats.

A social role of the CF value could be rooted in the potential phenotypic or genotypic quality information provided by the signal. Following Krebs & Singleton (1993), a body condition index (BCON) was calculated as the ratio between the actual body mass and the mass predicted by a ln—ln regression of body mass vs. FA for all individuals with available data (colonies 3, 5, 7 and 12). Because we were interested in the individual effect of body condition, the variation due to SEX and COLONY was removed by including these explanatory variables in the initial A N C O V A model. We then tested for a correla- tion between the residuals of the first GLM model for CF with BCON, for all data and within sex.

Table 2 Effect of POPULATION, COLONY and SEX on the resting frequency (CF) of Hipposideros ruber in the Gulf of Guinea.

Effect d.f. MS F P

POPULATION 3 896.301 49.74 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download