Temporal Patterns In The Emergence Behaviour Of Pipistrelle Bats

Transcript

Anim. Behav., 1995, 50, 1147-l 156 Temporal patterns in the emergence behaviour of pipistrelle bats, Pipistreffus pipistrellus, from maternity colonies are consistent with an anti-predator response J. R. SPEAKMAN, Department (Received final R. E. STONE Kz J. E. KERSLAKE of Zoology, University of Aberdeen 19 November 1993; initial acceptance 21 January acceptance 6 February 199.5; MS. number: 4530) 1994; Many species of bats emerge from maternity colonies in a non-random manner; however, quantifying the extent to which the emergence differs from random is complicated by the underlying variation in intensity of the emergence. In this paper an empirical method which overcomes these analytical difficulties is used to examine the functional nature of the pattern of bat emergence. The emergence behaviour from two large colonies of pipistrelle bats containing between 102 and 976 individual bats was monitored during the summers of 1992 and 1993. Both colonies had two primary exit holes with bats from one hole crossing the face of the second hole. Emergence events at the first hole triggered events at the second hole at both colonies demonstrating that the clumped emergence pattern was not a bottleneck artefact. The extent of clumping was greatest during the first half of the emergence, when light levels were high, suggesting an anti-predation function. This pattern was not a consequence of a higher emergence intensity early in the evening. 0 1995 The Association for the Study of Animal Behaviour Abstract. During the summer, female bats of many species gather together in colonies to gestate, give birth and rear their young to independence (Hill & Smith 1984). These ‘maternity’ colonies may comprise from as few as three or four individuals (e.g. Kunz 1982) to the enormous congregations of the Brazilian free-tailed bat, Tadarida brasiliensis, which may number up to 20 million (Davis et al. 1962). Each night the bats from a maternity colony normally emerge to feed and drink (Erkert 1982). (In this paper we use the term ‘emergence’ to refer to the movement from inside the colony to outside. This usage is independent of the number of bats involved.) When there are many bats in the colony the emergence is a conspicuous and highly predictable phenomenon. Such emergences often therefore attract predators which attack and feed on the emerging stream of bats (Gillette & Kimbourgh 1970). At intermediate numbers of bats in the colonies, between 25 and 2000 individuals, the emergence appears to consist of groups of bats emerging together, with longer gaps when few or no individuals come out. These groups have been called Correspondence:J. R. Speakman,Department of Zoology, University of Aberdeen, Aberdeen AB9 2TN, U.K. (email: [email protected]). 0003-3472/95/111147+10 $12.0010 ‘outbursts’ (Swift 1980) or ‘clusters’ of emerging bats (Bullock et al. 1987). One problem with this subjective interpretation of the behaviour, however, is that assignments of emerging animals into groups, clusters or outbursts may not reflect any objective reality in the actual emergence behaviour (Speakman 1993). That is bats emerging at random might be expected to vary in their interemergence times, and observers may subjectively class bats separated by small, but random, intervals as in a group, and longer random intervals as a gap between groups. To detect significant clustering in the emergence it is necessarytherefore to compare the expected inter-event interval distribution with that expected for a random emergence (Slater 1974; Sibly et al. 1990). A fundamental assumption underlying the statistical analyses that have been traditionally used to assessclustering in behaviour (log survivorship analysis: Slater 1974 and log frequency analysis: Sibly et al. 1990) and indeed more recent approaches to this type of analysis (Berdoy 1993), however, is violated in the emergence patterns of bats: and also incidentally in many other patterns of behaviour (Speakman et al. 1992). The intensity of the process generating the events is assumed to be constant. In bat emergences the intensity commonly follows a peaked function over time (e.g. 0 1995 The Association 1147 for the Study of Animal Behaviour 1148 Animal Behaviour, 50, 5 Speakman 1993), which may produce a spurious illusion of clustering in the traditional analyses, particularly if the emergence involves more than 100 individuals (Speakman et al. 1992). Formal demonstrations of clustering behaviour in bat emergences (e.g. Bullock et al. 1987) may thus be an artefact of violation of the constant intensity assumption, and there may thus be no objective phenomenon needing evaluation. An empirical methodology has been worked out to overcome this analytical problem (Speakman et al. 1992; Speakman 1993) and, when it was applied to colony emergence data, significant clustering was still detected (Speakman et al. 1992). Although the clustering phenomenon is not an artefact of the traditional statistics used in this type of analysis, it may still be an artefact of the movement of a large number of animals through a restricted exit hole. The presence of a restriction alone would be expected to generate a regular rather than a clumped or clustered emergence (Speakman et al. 1992); however, if this restriction was combined with a random variation in hesitancy of bats at the exit before emerging, animals with long hesitancies would hold up following bats behind them and this might produce clustering. This has been termed the ‘bottleneck hypothesis’ (Bullock et al. 1987). Alternatively the behaviour may have a more important biological function. First, since the emergences often attract predators the clustering pattern may represent a selfish anti-predator strategy (Hamilton 1964), with individuals emerging in groups to dilute their own probability of being predated. Second, emerging bats may group together as a feeding behaviour. This could be parasitic behaviour with individuals assessingthe successof other individuals and following them (thus emerging as a group) to maximize their own success(Ward & Zahavi 1973). Alternatively bats may derive significant mutual successif they feed in groups, for example by enhanced location of patchily distributed resources, and may thus emerge in clusters to synchronize themselves for subsequent foraging activity together (Wilkinson 1992). Attempts to separate these alternative hypotheses have so far proved difficult. Plastic and stuffed owls placed immediately outside colony exits have generally had only transient and insignificant effects on the extent of clustering (Speakman et al. 1992; Kalcounis & Brigham 1994). However, these manipulations are weak as the threat perceived by the bats was unknown, and in any case the major avian threat to bats is probably from diurnal rather than nocturnal predators (Speakman 1991). No experiments have yet been performed using live predators (either diurnal or nocturnal) placed near the colony exit. We aimed to assess whether the clustering previously observed in emergences of pipistrelle bats from maternity colonies was due to a bottleneck artefact. Second, we aimed to resolve whether clustering was more consistent with the predation or feeding hypotheses. To test between the alternative explanations we used two novel approaches. To establish if the emergences represented an artefact we located colonies where there were two exit holes close together. We reasoned that if emergence behaviour was simply an artefact of the numbers of bats moving through small spacesthen there would be no temporal concordance between events at the adjacent holes. However, both the feeding and predation hypotheses might be expected to lead to some extent of temporal concordance in clustering behaviour, particularly if bats from one hole regularly flew past the face of the second hole. Second, we reasoned that since diurnal predation is probably a much greater threat than nocturnal predation (Speakman 1991), if bats were emerging in clusters as an anti-predatory behaviour then more clustering would be evident in the first half of the emergence when it is lighter. METHODS Study Sites During the summer pipistrelle bats form maternity colonies, comprising approximately 98% females (Speakman et al. 1991), which vary in size from 10 to around 1500 individuals (e.g. Avery 1991). These colonies generally roost together in the roof space or cavity wall space of a building. In this paper we use the term colony to refer to a group of bats occupying a site and roost to refer to the specific location within that site where the bats routinely huddled together during the day. We studied two maternity colonies of pipistrelle bats in northeast Scotland (57”N). Both of the colonies were selectedfrom a total available pool of about 90 pipistrelle colonies in the local area (Speakman et al. 1991) because they had two adjacent emergence holes. The first colony was a Speakman Internal et al.: Emergence descent Figure 1. Schematic diagram of emergence at colony 1. The bats roosted in the apex of a dormer window and when emerging crawled down the beams from this roost to two emergence holes (ca 04 m). The two holes were located above windows and were about 1 m apart. Bats invariably flew to the west where the feeding site was located, thus bats emerging from hole A passed across the face of hole B. three-storey detached house, 25 km northwest of Aberdeen. The bats roosted in the apex of a dormer window at the west end of the house (Fig. 1). It was possible to get inside this roof and observe bats inside the apex, although for logistic reasons we could not do this at dusk when the bats were actually emerging. At dusk the bats moved down from this apex to two exit holes under the eaves at opposite ends of the dormer window. They did not need to fly from their habitual roosting location to the exit holes but needed only to crawl down the beams connecting the roost to the floor of the dormer. If bats were disturbed at the roost during the day this is commonly what they did. The internal distance between the normal roosting location of the bats and the exit holes was less than 1 m. The two exit holes were approximately 1 m apart. There were no other holes used by bats to exit or enter the colony. From inside the dormer the exit holes appeared to be about l-2 cm in diameter. It was unlikely that more than one bat could exit at one time and we never observed this to happen from outside. Bats inside the colony could move freely between the two exit holes. So once they crawled down from the roost to the exit hole have to crawl back up to the roost behaviour of buts 1149 down to the other one if they wished to change exit holes. If they crawled across the floor of the dormer, moving between the holes would probably have taken a bat about 5 s. If they crawled upside down along the inside of the dormer ceiling it would have taken them about 30-60 s to move between the holes. (Timings estimated on basis of observations of bats in the roost during the day.) The colony was about 1 km west of the river Don where most bats from this colony fed. Bats emerging from the colony invariably flew immediately eastwards across the face of the house. Consequently bats emerging from the most westerly hole (henceforth hole A) flew directly across the face of the more easterly hole (henceforth hole B; Fig. 1). The reverse was never observed to happen although very occasionally a bat from hole A exited westwards and did not cross the face of hole B. Since the windows that the two exit holes were above pointed slightly away from each other (Fig. 1) bats at hole A would have found it difficult to respond to the emergence of bats at hole B, but bats at hole B could easily respond to the bats from hole A as they crossed the exit hole. In 1992 there were only 1420 individual bats in this colony (J. E. Kerslake, personal observation). Observations were thus made only in 1993 when the colony contained considerably more bats (ca 290-970 individuals). Numbers of bats at this colony have varied between 15 and 1500 over the past 10 years for no apparent reason. The second colony was a single storey farm house about 30 km west of Aberdeen and 20 km south of the first colony. This colony contained 200-500 bats. Unlike the first colony the internal configuration of the second was unknown since it was not possible to get inside and see the bats. The bats emerged from underneath a gutter on the face of the house from three different holes. Hole A was almost in the centre of the house above the front door. Hole B was 2 m along the roof to the west of hole A and hole C was a further 3 m along the roof about 1 m from the west corner. The house faced directly onto, and was about 50 m away from, the river Feugh, along which most of the bats from this colony fed. Bats emerging from the colony generally flew towards the northeast across the face of the house, from holes A and B across hole C, although sometimes bats from any -of the holes would just fly directly out from the lony towards the river and would not therefore ss the face of other emergence holes. We 1150 Animal Behaviour, made observations at this colony in both 1992 and 1993. Both colonies were well established and have been occupied for at least 10-15 years, and probably for much longer. Emergences were observed from June through to September. During this period the bats gave birth (normally in early July) and the young flew during August and September. Variations in colony size within and between colonies therefore reflected both variations in the numbers of adult bats using the sites and also recruitment of young bats into the population. Independent of numbers of bats involved, the emergenceslasted approximately 1 h at both sites. The very first bats to emerge came out when it was still very light (mean=250 Ix, range 130-300) and when both potential predatory (e.g. sparrowhawks, Accipiter nisus) and potential aerial insectivorous competitive birds (e.g. swifts, Apus apus) were still active. The last bats to emerge came out when it was quite dark (2-5 lx). A single observer, or more rarely two observers, positioned outside the building, about 10 m away from the emergence holes, logged the emergence behaviour. We normally arrived at the colony about 60-90 mitt before sunset to make sure that all the emerging bats were recorded and also to settle in to the observation site before bats started emerging. In 1992 the emergence was logged by speaking each emergence event into a taperecorder which was subsequently played back and translated using a data logging program running on 486 PC (Viglen Ltd). In 1993, emergence events were recorded directly in the field using a portable laptop computer (T2000: Toshiba Ltd). The data logging program in both cases logged the time of each event with an identifier for the emergence hole each time a key was pressed. This allowed the data for each hole to be separated later during analysis. At the start and end of each emergence the air temperature (digital thermometer) and light levels (camera light meter: Minox Ltd) were also recorded. Light levels were later converted to lux using a calibration of the camera meter to a precision light meter (Gossen Mastersix). Data Analysis We sought to assessthe extent to which the emergence behaviour of bats at one hole might influence the emergence of bats at the other hole. 50, 5 In particular we were interested in whether bats that emerged from one hole (hole A at colony 1 and holes A/B at colony 2) and then flew across the face of the second hole (hole B at colony 1 and hole C at colony 2) triggered emergence events at that second hole. This temporal concordance between emergence at separate holes would indicate the emergence behaviour was not simply an artefact of large numbers of bats moving through a small space. We termed the presence of this temporal concordance in emergence events across two holes ‘co-clustering’. To test for co-clustering in the emergence data, we analysed the extent of clustering in the entire emergence pooling the data from the two exit holes, and using the trimming procedure outlined by Speakman et al. (1992) and then an optimized chi-squared test using the analysis software ‘CLUSTAN’ (Speakman 1993). The data were then separated into two streams reflecting emergences at each of the separate exit holes. We analysed the data separately for each hole to establish that emergence at each hole was independently clustered. If the emergence was co-clustered then the extent of clustering in the entire emergence would be greater than expected from a random juxtaposition of the two independently clustered event streams. To test this possibility we took the emergence data from the second hole, across which bats from the first hole passed as they flew to the foraging site (hole B at colony 1 and hole C at colony 2) and randomly rearranged the inter-emergence intervals. This randomization was performed using the random number generator and sorting functions (RAND and SORT respectively) in the statistical package MINITAB (Ryan et al. 1985). The randomization process retained the clustering structure of the emergence at the second hole but destroyed any temporal concordance between the emergence streams from the two holes. We then reconstructed a hypothetical total emergence using this randomized sequence for the second hole. This randomized hypothetical emergence sequencewas then analysed in the same way as the actual total emergence. The randomization and generation of hypothetical emergence sequencesin which the temporal concordance of events at the two holes was destroyed were each performed 100 times for each logged emergence. In this way, for each emergence that was logged, we generated a Speakman et al.: Emergence behaviour of bats 1151 Actual emergence Chi-squared value Figure 2. Distribution of 100 randomized chi-squared values generated by a randomization procedure to simulate the expected extent of clustering if there was no co-clustering between exit holes, from an emergence of pipistrelle bats from colony 1. The actual emergence chi-squared value from this emergence is shown. distribution of chi-squared values for the expected The extent of clustering in each half of the emerextent of clustering in the total emergence if there gence was then measured using the CLUSTAN was no co-clustering between the separate holes. software package. Previous studies have suggested The distribution of chi-squared values for the that risk of predation is about 100-1000 times 100 randomized emergence streams was approxi- greater in daylight than at night (Speakman 1991, mately normal. We tested for normality in these in press). Consequently, if the function of clusterdistributions by comparing the observed distri- ing was an anti-predator responsewe might expect butions to generated normal distributions centred significantly more clustering in the first half of the on the same means and with the same standard emergence. We compared the extent of clustering deviations using the RAND and NORM com- in the two halves of the emergences at one site mands in MINITAB. In all but two cases the using a paired t-test. distributions were not significantly different from normal (Kolmogorov-Smirnov test: Sokal & RESULTS Rohlf 1981). We therefore tested the extent of co-clustering in the actual emergence by comparing the extent of clustering (chi-squared value We logged 3 1 complete emergencesacross the two from CLUSTAN) to the mean and standard colonies in 1992 and 1993 (15 at colony 1 in 1993 deviation chi-squared across the 100 randomized and 11 at colony 2 in 1992, five in 1993). The total hypothetical emergences using a z-test (Sokal & numbers of exiting bats varied between 293 and Rohlf 1981). A one-tailed test was employed as we 976 individuals at colony 1 and between 102 and predicted a priori that a positive deviation would 698 at colony 2. be expected for co-clustering. To assesswhether clustering in the emergence Co-clustering behaviour was consistent with an anti-predator Figure 2 shows a typical distribution of the response we divided the total emergence stream (pooled acrossboth emergence holes), and divided chi-squared statistic reflecting the amount of it in half at the emergence time of the median bat. clustering in the emergence behaviour for the 1152 Animal Behaviour, Table I. The number of bats (N) emerging at two 50, 5 were emerging from hole C but a significant coloniesof pipistrellebats in northeast Scotlandwith a number (D-40) were also emerging from holes A chi-squaredstatisticrepresentingthe extent of clustering and B (combined). Unfortunately only four of the for the total emergenceacrosstwo exit holes observed nights met these criteria. At colony 2 Colony N Z* we recorded significant co-clustering on three of pt x2 four comparisons where the analysis was feasible 1 293 112.3 5.19