Discriminating predation attempt outcomes during natural foraging using the post-buzz pause in Japanese large-footed bat Myotis macrodactylus

Bats emit a series of echolocation calls with an increasing repetition rate (the terminal buzz), when attempting to capture prey. This is often used as an acoustic indicator of prey-capture attempts. However, because it is directly linked to foraging efficiency, predation success is a more useful measure than predation attempts in ecological research. The characteristics of echolocation calls that consistently signify predation success across different situations have not been identified. Due to additional influencing factors, identification of these characteristics is particularly challenging for wild bats foraging in their natural environment compared to those in flight chambers. This study documented the natural foraging behavior of wild Japanese large-footed bat Myotis macrodactylus using synchronized acoustic and video recordings. From the video recordings, we could assign 137 attacks to three outcome categories: prey captured (51.8%), prey dropped (29.2%), and failed attempt (19%). Based on previous indications from laboratory studies that the length of the silent interval following the terminal buzz (post-buzz pause) might reflect the prey capture outcome, we compared post-buzz pause durations among categories of attack outcomes. The post-buzz pause was longest in the case of successful capture, suggesting that the length of the post-buzz pause is a useful acoustic indicator of predation success during natural foraging in M. macrodactylus. Our finding will advance the study of bat foraging behavior using acoustic data, including estimations of foraging efficiency and analyses of feeding habitat quality. Summary statement We investigated the natural foraging behavior of wild Myotis macrodactylus and found that the length of the post-buzz pause is a useful acoustic indicator of predation success.

6 al., 2011). The distance between the central microphone, M1, and each of the three equally spaced 1 1 5 outer microphones was 0.9 m (Fig. 2B). The four units in the microphone arrays (arrays 1-4) were 1 1 6 arranged such that echolocation calls emitted from anywhere around the pond area could be recorded. 1 1 7 The echolocation signals recorded by the microphones were amplified and band-pass 1 1 8 filtered (10~250 kHz) using a custom-designed electronic circuit, and then digitized with 16-bit 1 1 9 precision at a sampling rate of 500 kHz using a high-speed data acquisition card (PXIe-6358; 1 2 0 National Instruments, Austin, TX, USA). The frequency response of the microphones was flat (±3 1 2 1 dB), and ranged between 10 and 250 kHz. The output signals were synchronously stored using a 1 2 2 personal computer via a custom program created using LabVIEW 2011 (National Instruments). 1 2 3 Recordings were saved as files every 10 minutes, and recording was stopped when the batteries ran 1 2 4 out. 1 2 5 Sound data from the central microphone in each of the four microphone arrays (ch 1, 5, 9, 1 2 6 13) were analyzed. We used Cool Edit 2000 (Syntrillium Software Corporation, Phoenix, AZ, USA) 1 2 7 to display the spectrograms of the sounds (128-point FFT, Han window with overlap for N-1) and 1 2 8 extracted the terminal buzz signals with the clearest spectrograms for analysis. When the signal-to-1 2 9 noise ratio of a pulse was poor and extraction of the sound data was difficult, the data from the other 1 3 0 channels were checked to determine the presence or absence of a terminal buzz. The post-buzz pause 1 3 1 was calculated as the time between the last sound in the terminal buzz and the start of the next search 1 3 2 pulse, as measured from the spectrogram images (Fig. 1A).

3 3
Video recording and analysis 1 3 4 The foraging behavior of the bats at the pond was recorded using high-speed cameras (LT 1 3 5 Recorder Pro (ver. 1.04; DITECT, Tokyo, Japan), in synchronization with the above-described 1 3 6 sound recordings. The observation area was illuminated by infrared floodlights (LIR-CS88; IR LAB, 1 3 Shenzhen, China) and the frame rate of the camera was set to 60 fps. An analog on/off control signal 1 3 8 generated by a custom-made electrical circuit triggered video recordings, so that the video and sound 1 3 9 data could be synchronously recorded and stored on the PC. The video recording was stopped when 1 4 0 the hard drive of the computer was full (after approximately 30 minutes per measurement day).

4 1
Video images were analyzed visually using Dipp-Image Viewer (version 1.22; DITECT). 1 4 2 In the first step, we identified scenes that showed a bat attacking prey, i.e., cases in which the tail 1 4 3 membrane and hindfeet of a bat touched the water surface. These scenes were classified as "catch" 1 4 4 or "failed", and the "catch" group was further classified as "captured" (successful predation) or 1 4 5 "dropped" (Fig. 3). Scenes were classified as "captured" when the bat caught the prey near the water 1 4 6 surface with its feet or tail membrane and carried it away. Scenes where a bat caught prey but then 1 4 7 dropped it were classified as "dropped". The "failed" category contained scenes in which the 1 4 8 presence of the prey on the water surface was confirmed after the bat had attacked. sequence, and therefore from the same bat, we included "bat-ID" as a random effect in all models. 1 5 6 Also, to account for any differences between recording events, we included the date of the recording 1 5 7 as a random effect in all models. Hereby, the effect of bat-ID was nested within the date factor. The 1 5 8 9 from the spectrograms with a good signal-to-noise ratio (capture: n = 37, drop: n = 33, failed: n = 17). 1 8 3 Based on the minimum AIC value, the model containing only the attack outcome factor was the best 1 8 4 (Table 2). After graphically examining the model residuals, the fit was determined to be satisfactory. 1 8 5 The best model explained significantly more variance than the null model (χ 2 = 162.14, degrees of 1 8 6 freedom [df] = 2, P < 0.001), and the attack outcome factor was significant (χ 2 type-II-Wald = 158.1, df = 1 8 7 2, P < 0.001). The post-buzz pause was longest in cases of successful capture, with a mean ± SE 1 8 8 value of 200 ± 11.14 ms (χ 2 type-II-Wald test: capture vs. failed ratio = 0.573 ± 0.03, df = 82, P < 0.001, 1 8 9 capture vs. drop ratio = 0.767 ± 0.03, df = 82. P < 0.001, Fig. 4). The post-buzz pause was shortest in 1 9 0 cases of failed capture, with a mean value of 114 ± 7.05 ms (χ 2 type-II-Wald test: drop vs. failed ratio = 1 9 1 1.340 ± 0.05, df = 82, P < 0.001), while the mean post-buzz pause in cases of dropped prey had an 1 9 2 intermediate value of 153 ± 8.64 ms. 1 9 3 We observed additional interesting behaviors in the video recordings. For example, the bat 1 9 4 attacked small branches ("Attack branch" in Movie 2) or the same prey again after initially failing to 1 9 5 capture it ("Repeated attack" in Movie 2). In addition, in cases of dropped prey, the bats usually 1 9 6 dropped the item immediately after capture. However, there were a few cases where the bats 1 9 7 dropped objects after holding them for a longer period ("Throw away" in Movie 2). 1 0 , 1999;Surlykke et al., 2003;Übernickel et al., 2013). Laboratory experiments on foraging 2 0 6 behavior have indicated that the presence or absence of predation affects the post-buzz pause 2 0 7 (Acharya and Fenton, 1992;Britton and Jones, 1999;Surlykke et al., 2003;Übernickel et al., 2013).

0 8
In contrast, no clear evidence for such a relationship could be derived from observations of wild bats 2 0 9 during natural foraging (Britton and Jones, 1999;Surlykke et al., 2003). This was due to poor-2 1 0 quality acoustic data and the tendency for post-buzz pauses to be relatively short as a result of 2 1 1 adaptation by the bats to the complex natural environment. To the best of our knowledge, this study 2 1 2 is the first to show that the length of the post-buzz pause can be used to measure successful predation 2 1 3 by wild bats during natural foraging. In the present study, M. macrodactylus bats dropped their prey during 30% of all recorded 2 1 7 attacks (40 drops out of 137 attacks). Therefore, these bats do not appear highly capable of 2 1 8 discriminating their prey, but rather make their prey selection after capture. This may explain the 2 1 9 intermediate values of the post-buzz pause observed in this study.

0
Previous studies on prey selection in bats that predominantly hunt via trawling, like M. 2 2 1 macrodactylus, have also suggested that bats have relatively weak target discrimination ability.

2
Indeed, bats have been observed to sometimes attack objects instead of prey (Barclay and Brigham, 2 2 3 1994; Kalko and Schnitzler, 1989). For example, Myotis lufigusus and M. yumanensis did not appear 2 2 4 to discriminate among targets, and attacked inedible targets (beetles and leaves) as well as edible 2 2 5 prey of the same size (moths) during natural foraging (Barclay and Brigham, 1994). Other trawling 2 2 6 Myotis species, such as M. dasycneme, M. daubentonii, and M. capaccinii, repeatedly attempted to 2 2 7 capture inedible dummy targets placed on artificial surfaces that mimicked the reverberatory 1 1 properties of water (Siemers et al., 2001). Therefore, this might represent a general prey selection 2 2 9 behavior in trawling bats. 2 3 0 On the other hand, bats might drop not only inedible targets, but also edible prey 2 3 1 unintentionally. For instance, in the "repeated attack" shown in Movie 2 the bat most likely dropped 2 3 2 an edible prey item because it recaptured that item after dropping it. However, since this type of 2 3 3 behavior was captured rarely by our cameras, there is a limitation to discriminate between these two 2 3 4 types of behaviors in this study. In the future, knowing what the bats have caught would help 2 3 5 distinguishing between drops of inedible or edible targets.

3 6
The post-buzz pause was also shown to be affected by prey size in a laboratory study with 2 3 7 Pipistrellus pygmaeus (Surlykke et al., 2003). However, the authors reported that the prey size did 2 3 8 not have a significant effect on the post-buzz pause in wild bats, according to acoustic recordings 2 3 9 taken during natural foraging.

4 0
In summary, we found a clear relationship between the post-buzz pause and predation 2 4 1 success in naturally foraging M. macrodactylus. However, further investigation regarding this 2 4 2 relationship, including influencing factors such as prey type and size, will be needed to develop a 2 4 3 reliable acoustic indicator of predation success in wild bats. To elucidate what insights could be gained by investigating post-buzz pauses during 2 4 7 natural foraging, we recorded the foraging behavior of individual M. macrodactylus bats using four 2 4 8 microphone arrays surrounding the pond, in the same experimental setting, for about 100 minutes 2 4 9 M i z u g u c h i , e t a l . 1