PreprintPDF Available

Echolocating bats prefer a high risk-high gain foraging strategy to increase prey profitability

Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Most bats catch nocturnal prey during active flight guided by echolocation but some species depart from this ancestral behaviour to capture ground prey using passive listening. Here, we explore the costs and benefits of these hunting transitions by combining high-resolution biologging data and DNA metabarcoding to quantify the relative contributions of aerial and ground prey to the total food intake of wild greater mouse-eared bats. We show that these bats use both foraging strategies with similar average nightly captures of 25 small, aerial insects and 30 large, ground-dwelling insects per bat, but with higher capture success in air (78 % in air vs 30 % on ground). However, owing to the 3 to 20 times heavier ground prey, 85 % of the estimated nightly food acquisition comes from ground prey despite the 2.5 times higher failure rates. Further, we find that most bats use the same foraging strategy on a given night suggesting that bats adapt their hunting behaviour to weather and ground conditions. We conclude that prey switching matched to environmental dynamics plays a key role in covering the energy intake even in specialised predators.
Content may be subject to copyright.
Echolocating bats prefer a high risk-high gain foraging strategy to increase
prey profitability
Laura Stidsholt1*, Antoniya Hubancheva2,4, Stefan Greif2,3, Holger R. Goerlitz2, Mark
Johnson1, Yossi Yovel3, and Peter T. Madsen1
1 Zoophysiology, Department of Biology, Aarhus University, Aarhus, Denmark.
2 Acoustic and Functional Ecology, Max Planck Institute for Biological Intelligence,
Seewiesen, Germany.
3 Department of Zoology, Tel Aviv University, Tel Aviv, Israel
4 Department of Animal Diversity and Resources, Institute of Biodiversity and Ecosystem
Research, Bulgarian Academy of Sciences, Sofia, Bulgaria
ORCID: LS: 0000-0002-2187-7835, AH: 0000-0001-8362-1301, SG: 0000-0002-0139-2435, HRG:
0000-0002-9677-8073, MJ: 0000-0001-8424-3197, YY: 0000-0001-5429-9245, PTM: 0000-0002-
Key words
Biologging, bio-acoustics, foraging strategies, foraging behaviour, prey profitability, prey
switching, high risk-high gain strategy.
Abstract 1
Most bats catch nocturnal prey during active flight guided by echolocation but some species 2
depart from this ancestral behaviour to capture ground prey using passive listening. Here, we 3
explore the costs and benefits of these hunting transitions by combining high-resolution 4
biologging data and DNA metabarcoding to quantify the relative contributions of aerial and 5
ground prey to the total food intake of wild greater mouse-eared bats. We show that these bats 6
use both foraging strategies with similar average nightly captures of 25 small, aerial insects 7
and 30 large, ground-dwelling insects per bat, but with higher capture success in air (78 % in 8
air vs 30 % on ground). However, owing to the 3 to 20 times heavier ground prey, 85 % of the 9
estimated nightly food acquisition comes from ground prey despite the 2.5 times higher failure 10
rates. Further, we find that most bats use the same foraging strategy on a given night suggesting 11
that bats adapt their hunting behaviour to weather and ground conditions. We conclude that 12
prey switching matched to environmental dynamics plays a key role in covering the energy 13
intake even in specialised predators. 14
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Introduction 15
Bats are widespread and abundant predators that serve important roles in ecosystems across 16
the globe (Kunz et al., 2011).Their evolutionary success is due to the unique combination of 17
echolocation and powered flight (Teeling et al., 2005) allowing them to avoid visual predators 18
by feeding at night, thereby gaining unfettered access to food sources that include insects, small 19
vertebrates, fruit, nectar, pollen and blood(Simmons, 2005). Within the mosaic of foraging 20
niches they exploit, echolocating bats are categorised into three main foraging strategies: the 21
ancestral mode of capturing prey on the wing (hawking), and the derived modes of trawling 22
prey from water surfaces or gleaning prey, nectar or fruit from ground and trees(Schnitzler and 23
Kalko, 2001). To aid these specialised hunting strategies, each guild of bats has evolved 24
specific adaptations in echolocation signals, auditory systems, morphology, and flight 25
mechanics (Norberg and Rayner, 1987; Fenton, 1990; Schnitzler and Kalko, 2001). Despite 26
such specialism, recent research has shown that foraging style is not monotypic within species: 27
gleaning bats occasionally capture aerial prey (Bell, 1982; Fenton, 1990; Ratcliffe and Dawson, 28
2003; Ratcliffe, Fenton and Shettleworth, 2006; Hackett, Korine and Holderied, 2014), while 29
insect-gleaning bats may seasonally target nectar or fruit(Aliperti et al., 2017) or vice versa 30
(Herrera et al., 2001). These changes in foraging style presumably track the relative abundance 31
of preferred versus alternative food sources, broadening the ecological roles of bats and 32
providing a degree of resilience in the face of changing resources. However, owing to the 33
complexity of studying detailed hunting behaviours in the wild, it is not clear why or when 34
specialised bats switch foraging strategies. This led us to ask whether bats adapt their hunting 35
strategies continuously to maintain net intake or if switching is the last resort when preferred 36
prey are unavailable. To address that, we used miniaturized biologging devices to track the 37
hunting behaviour of greater mouse-eared bats. This species primarily captures ground-38
dwelling arthropods by passively listening for their movements(Arlettaz, 1996), and is 39
therefore specialised for gleaning: broad, short wings enable them take-off from the ground, 40
and weak echolocation calls avoid alerting prey while also allowing the bat to hear rustling 41
sounds of surface-dwelling prey. Nonetheless, like many other gleaning bats, greater mouse-42
eared bats have maintained the ancestral ability to use echolocation to capture aerial prey on 43
the wing, which requires that they switch to intense calls to detect small prey, and maintain the 44
capability to manoeuvre fast in 3D space to track evasive prey. While call intensity can be 45
adjusted to fit different strategies, the morphological and anatomical specialisations for ground 46
gleaning must affect the efficiency of these bats as aerial hawkers. We therefore predicted that 47
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
bats would prefer gleaning whenever it was profitable, and would only switch to aerial hunting 48
when environmental conditions led to poor prey biomass per foraging time when gleaning. To 49
test this prediction, we used sound and movement tags (N = 34 bats) (Stidsholt et al., 2018) to 50
record the bats’ echolocation behaviour, 3D movement patterns, GPS locations (N = 7 of 34 51
bats) and mastication sounds after prey captures to quantify foraging success. We augmented 52
these data with DNA metabarcoding of faeces from co-dwelling con-specifics (N = 54 bats) to 53
identify prey species and sizes. This combined biologging and DNA metabarcoding approach 54
allows us to estimate prey profitability across foraging strategies and habitats to evaluate the 55
drivers of prey switching in wild bats. 56
Results 58
To categorise the foraging strategies used by wild bats, we analysed one night of sound and 59
movement data from each of 34 female, greater mouse-eared bats (Myotis myotis). All bats 60
commuted from the colony or release site to one or several foraging grounds before returning 61
back to the roost before dawn (Fig. 1ABC). Since we used two different types of biologging 62
devices with different sensors, foraging bouts were defined either as intervals of high variation 63
in heading, or as intervals of area-restricted search for tags including GPS (N = 7 bats). A total 64
of 3917 attacks on prey were recorded with most bats capturing prey both on the ground by 65
passively gleaning prey, and by pursuing prey mid-air by aerial hawking (Fig. 1DE). However, 66
four bats exclusively gleaned, while two bats only hawked (Fig. 1DE). The dominant foraging 67
strategy used per bat per night seemed to be affected by the night of tagging indicating that bats 68
tagged on the same nights choose the same foraging strategy (Fig. 1D & Fig. S1, N = 10 nights, 69
1 to 9 bats tagged per night; LMM; testing if the ratio between ground:aerial captures was 70
explained by night of tagging, p=0.001). 71
The bats attacked food more often on the ground (mean: 80 attacks per individual per night, 72
quartiles: 26110) than in the air (mean: 35 attacks, quartiles: 670; LMM, p=0, Table S3, Fig. 73
1D), but the proportion of attacks that were successful (i.e. success ratio) based on audible 74
mastication sounds following prey captures were more than double in air (79 %, quartiles: 7175
88) than on ground (31 %, quartiles: 25–40; Table S4, Fig. 1F). This led to on average 25 76
(quartiles: 10 to 33) aerial and 30 (quartiles: 5 to 53) ground insects caught per bat per night 77
(Fig. 1D). Prey attack rates were substantially higher for ground foraging versus aerial foraging 78
(ground: 1.73 (quartiles: 1.7-1.75) prey attacks/minute vs aerial: 1.17 (quartiles: 1.14-1.21)) 79
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
(Fig. 1G). Thus, bats caught prey much more reliably in the air, but attacked ground prey more 80
often and devoted more foraging time to ground gleaning. 81
Fig. 1: Greater mouse-eared bats tagged on different nights show wide variation in 84
foraging strategy and success. A-C: The jerk (differential of on-animal recorded acceleration) 85
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
reveals the overall movement of the bat by showing periods of no movement (rest) and strong 86
movement (flight) for three different bats (summed values for Bat ID 25,5 and 30 as depicted 87
in panel D) and two different travel modes (commuting (dark grey) vs foraging (light grey)). 88
We marked all prey attacks as either hawking (blue diamonds) or gleaning (green circles) by 89
visual and auditory inspection of the sound and movement data. Prey captures were classified 90
by audible mastication sounds as successful (black edge) or failures (pink edge). The bats 91
exemplified here either primarily gleaned (A), primarily hawked (B) or used both strategies in 92
alternating bouts (C). D-E) Successful (D) and unsuccessful (E) prey attacks of all bats (N = 93
34) grouped according to night of tagging for aerial hawking (blue) and gleaning (green). Stars 94
mark the bats equipped with GPS tags; A, B, C mark the bats depicted in panels A-C. F-G) The 95
success ratio (F) reveals the percentage of all attacks that were successful per bat per night 96
(dots), while attack rates (G) reveal the number of foraging attacks per minute for each bat per 97
night (dots) with more than one prey attack per foraging strategy for aerial hawking (blue) and 98
gleaning (green) along with kernel densities and boxplots. 99
We next used GPS tracks from seven bats to investigate the behavioural and ecological factors 101
that influence foraging success. Specifically, we tested if movement style (i.e., commuting vs 102
actively searching for prey defined via the Lavielle method(Hurme et al., 2019)) and habitat 103
(i.e. forest vs open fields) affected the success ratio and prey attacks. The GPS tagged bats also 104
predominantly caught prey in separate foraging bouts that were each dedicated to either 105
hawking (Fig. 2A, blue diamonds) or gleaning (Fig. 2A, green circles). However, almost half 106
of all aerial prey were captured during commuting (47 % of total aerial captures; Fig. S2). 107
When gleaning, the bats attacked the same total number of prey in forest and open field habitats 108
(field: 207 vs forest: 221 attack in total, Fig. 2C), but with more attacks per bout when gleaning 109
above fields (field: 25 vs forest: 14 attacks/bout). Moreover, gleaning in open fields was twice 110
as successful than in forest habitats (success ratio of 48 % per foraging bout with more than 2 111
prey attacks in open fields vs 12 % in the forest, Fig. 2E, Table S6). In contrast, when capturing 112
insects in air, the attack rates and success ratios were consistently high and unaffected by 113
habitat (Fig. 2DF, LMM, Table S5). 114
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Fig. 2: Habitat influences foraging success of greater mouse-eared only when gleaning. 116
A) Tracks of seven bats with GPS tags released either at the cave (red star) or at a location 117
nearby (white star) and their foraging behaviour: Gleaning (green circles) and hawking (blue 118
diamonds) attacks with success (yellow edge) or failure (black edge). B) The bats were tracked 119
in North-Eastern Bulgaria (white square). C-F: Total prey attacks (CD) and success ratios per 120
foraging bout (EF), for both habitats: open field (yellow; G) and forest (magenta; H). Each data 121
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
point corresponds to one foraging bout and is sized according to the number of attacks in the 122
bout. G-H: The two main foraging habitats of greater mouse-eared bats: open fields (G) and 123
the open spaces below the canopy in forests (H). 124
Fig. 3: Ground prey is larger than aerial prey and sufficient to offset the lower foraging 126
success ratios of gleaning. 127
A-B: DNA metabarcoding of faeces from 54 greater mouse-eared bats (48 females, 6 males). 128
Insects were categorized as either ground (green) or aerial (blue). The few species (N = 5) that 129
are both aerial and ground were omitted from the analysis. Distribution of the targeted prey 130
orders depicted as OTU (Operational Taxonomic Units) between ground (~40%) and aerial 131
(~60%) niches (A) and across taxonomical units in the Arthropoda (B). C-F): Prey properties 132
and profitability during gleaning (ground prey, green) and aerial hawking (aerial prey, blue), 133
with kernel densities and boxplots. C) Body lengths of the prey sorted by foraging style. D) 134
Number of mastication sounds identified after each prey capture by an automatic detector (N 135
= 244 ground captures and 336 aerial captures across 10 bats). E) Dry prey body masses of 136
each prey type identified for gleaning via DNA metabarcoding (green circles) (DNA 137
metabarcoding was used as the reference prey body mass for ground captures), and for aerial 138
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
prey by mastication analysis (blue triangles) and DNA metabarcoding (blue circles). F) Prey 139
profitability of gleaning or hawking prey based on prey body masses from mastication analysis 140
(triangles) or DNA metabarcoding (circles) combined with observed success ratios, handling 141
and search times. The data are plotted for bootstrapped data (N=70 random data points) due to 142
varying sample sizes of each parameter. G) Total caloric intake per night per bat calculated by 143
multiplying the caloric intake per prey with the number of successful gleaning and hawking 144
prey captures (Fig. 1), and compared to the field metabolic rate of a 30 g bat estimated from 145
the literature (orange). 146
To estimate prey types and sizes, we first performed DNA metabarcoding analysis on the faeces 148
of 54 untagged greater mouse-eared bats from the same colony caught in the morning upon 149
returning from the foraging grounds. The bats target a wide range of prey species spanning 155 150
OUT (Operational Taxonomic Units) (Fig. 3AB), of which ~60 % occupy aerial niches (36 151
families) and ~40 % occupy ground niches (23 families; Fig. 3A). We measured the lengths of 152
representatives of each species from online photo databases covering the same region in 153
Bulgaria. Ground prey were 2.5x longer than aerial prey (20 mm (quartiles: 14–28 mm) vs 7 154
mm (quartiles 4–9 mm), Fig. 3C). We used length-weight regressions(Straus and Avilés, 2018) 155
for each group to convert body-length to body mass for each prey type. For this analysis, we 156
used the weighted average of the two most numerous prey types for aerial and ground 157
regression values. Under these assumptions, estimated dry body masses of ground prey were 158
~20 times heavier than aerial prey (means: 67.5 mg (quartiles: 29.7–146.6 mg) vs 3.0 mg 159
(quartiles: 0.7–5.7 mg, Fig. 3E, green and blue circles). 160
Since DNA metabarcoding does not provide the exact proportion of caught prey items and 161
species, and thus does not allow to calculate the size distribution of caught prey, we performed 162
an analysis of the mastication sounds as an additional proxy for prey size. Greater mouse ear 163
bats chew all prey while flying irrespective of how they are caught and take longer to masticate 164
larger prey (verified in laboratory feeding experiments, Fig. S9). We used body length to body 165
mass conversions from the DNA metabarcoding of ground prey as the reference prey body 166
mass. We then estimated aerial prey body masses from the difference in chewing durations 167
between ground and aerial prey. Bats chew longer on ground prey than aerial prey, indicated 168
by the ~3x more mastication sounds detected after each gleaning capture (75, quartiles: 38.5-169
111.5, Fig. 3D) compared to aerial hawking (23, quartiles: 15- 55). By applying this ratio to 170
the body mass estimations of gleaning prey, we estimated aerial prey body mass of 21.2 mg on 171
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
average (quartiles: 9.3-45.9) (Fig. 3E, blue triangles). Thus, in the following, we use both a 172
lower and a higher estimate of aerial prey body masses of 3.0 mg (from DNA metabarcoding), 173
and 21.2 mg (from masticating sounds) (Fig. 3E). Taking successful prey captures into account 174
and the weighted average of the caloric values of the two most numerous prey types for aerial 175
and ground (25.4 kJ/g dry mass of ground prey and 21.3 kJ/g dry mass of aerial prey), the bats 176
ingested an average of 60.2 kJ/night/bat (quartiles: 32.1-84.8) based on the lower aerial prey 177
body mass estimates (Fig. 3G, solid grey line), and 74.9 kJ/night/bat (quartiles: 55.2-95.7) 178
based on the higher estimates (Fig. 3G, shaded grey). Using energy assimilation rates of 50-179
82% in bats(Kurta et al., 1989; Straus and Avilés, 2018), the bats obtained on average between 180
30-61 kJ/night per bat (Fig. 3G). 181
The profitability of prey caught by gleaning or hawking for all bats (N = 34) was quantified by 182
combining success ratios (Fig. 1F), search and handling times (Fig. S4A) with lower (Fig. 3E, 183
circles) and higher estimates of prey body masses (Fig. 3E, triangles). The gleaning foraging 184
strategy yielded a profitability of 7.4 J/sec (quartiles: 5.6-8.1 J/sec, Fig. 3F green), while 185
hawking resulted in a lower estimate of 0.5 J/sec (quartiles: 0.4-0.54, Fig. 3F, blue circles) and 186
a higher estimate of 3.3 J/sec (quartiles: 2.8-3.8, Fig. 3F, blue triangles). Prey profitability when 187
gleaning is thus 2.3-14 times higher than when aerial hawking. 188
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Discussion 190
The small size and high metabolic rate of bats, coupled with a costly locomotion mode, require 191
an elevated and constant input of calories from foraging(Kleiber, 1947). This, in turn, calls for 192
either stable and narrow food niches or adaptive hunting behaviours that track habitat 193
dynamics. Here, we used biologging and metabarcoding to explore how greater mouse-eared 194
bats chose between two different foraging strategies to cover their energy intake, and how 195
strategy switching is adapted to habitat. 196
Greater mouse-eared bats are more successful when hawking despite being gleaning 197
specialists 198
Since gleaning requires passive listening, while aerial hunting requires vocalising(Arlettaz, 199
1996), we first hypothesized that the two strategies were mutually exclusive and would yield 200
different prey attack rates and success ratios. Indeed, our data show that the tagged bats used 201
both foraging strategies to capture food, but that foraging bouts were dedicated exclusively to 202
either gleaning or aerial hawking (Fig. 1 & Fig. S11). Nonetheless, averaging over ten nights 203
and individuals, tagged bats caught a mean of 25 insects in air and 30 insects on the ground 204
during a night of foraging (Fig. 1), indicating a reliance on both food sources. 205
These estimated feeding intakes are slightly higher than the total prey captures of the similar-206
sized Rhinopoma microphyllum, estimated from buzz counts (Cvikel et al., 2015), but well 207
below previous indirect estimates of feeding intakes in a smaller (7-11g) species (M. 208
daubentonii) that were extrapolated to suggests that they should capture thousands of tiny 209
insects per night. The discrepancy between the total prey captures could potentially be a 210
consequence of the tag, or the tagging process. However, the extra weight of tags (~3-4 211
g)(Portugal and White, 2018; Kline, Ripperger and Carter, 2021) did not appear to strongly 212
impact the ability of bats to capture food since i) both tagged and un-tagged trained bats quickly 213
learned to intercept aerial and ground prey in the lab with similar success ratios as in the wild 214
(Table S2, Video S1-2), and ii) the wild tagged bats spent the same amount of time on foraging 215
outside the colony as bats equipped with lighter (0.4) telemetry transmitters (Egert-Berg, 216
Hurme, Greif, Wilkinson, et al., 2018). 217
Given our measured total prey captures and prey sizes, wild greater mouse eared bats in our 218
study assimilated an average of 61.4 kJ/night per bat (82 % assimilation). This is higher than 219
the estimated field metabolic rate (FMR) of the similar-sized female lesser long-nosed bats 220
(Leptonycteris yerbabuenae) of 40 kJ/day, but close to allometric scaling of the FMR of a 30 221
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
g bat based on heart rate measurements from wild-tagged 18 g Uroderma bilobatum (FMRM.myo 222
= FMRU.bil * (30g/18g)0.7 = 65.5 kJ/day). Thus, despite making less than 100 prey captures per 223
night, the estimated food intake of the tagged bats matches their predicted FMR. The 224
discrepancy between the predicted intake of thousands of insects per night from a smaller bat 225
species, and the measured total prey captures in our study is therefore more likely to relate to 226
the dramatic difference in prey sizes between the prey species rather than a reduced foraging 227
effort due to tagging effects. The bats in our study on average reached their predicted energetic 228
requirement in a full night of foraging, but only just so, indicating that they may have little 229
scope to compensate for changes in their environment. Since the bats fly out just after sunset 230
and return early in the morning, they are vulnerable to any disturbance or change in habitat 231
quality that reduces their foraging intake. Moreover, despite selecting only heavy, post-232
lactating females within the same colony, there was wide individual variation in hunting. 233
Tagged bats attacked from 48 to 280 prey during a night of foraging (Fig. 1-2), demonstrating 234
that continuous recordings from the same individuals are important to quantify the energy 235
budgets of wild animals. 236
Even though the tagged bats captured the same number of prey by aerial hunting as by 237
gleaning, we expected that their sensorimotor adaptation to gleaning would come at the cost of 238
a poorer ability to capture aerial prey during hawking. Counter to our hypothesis, we found 239
that hawking bats were highly successful (80%) compared to when gleaning (30%) (Fig. 1F). 240
Despite that greater mouse-eared bats are gleaning specialists, we find that they have success 241
ratios when hawking that are on par with observations in the wild for hawking bats (Rydell, 242
McNeill and Eklöf, 2002). These high success ratios may be facilitated by superfast 243
sensorimotor responses to guide echo-based capture. Such superfast movements may not 244
benefit a gleaning strategy to the same extent since carabid beetles can seek refuge under leaves 245
or twigs if the attack of the bat is not perfectly aimed on the prey. In such cases, the bats must 246
rely on tactile and olfactory cues, and their poorer ground locomotion to find the prey(Kolb, 247
1958), which may explain the low success rate of gleaning (Table S2). 248
Greater mouse-eared bats prefer larger ground-dwelling prey over aerial prey despite 249
low success ratios 250
Tagged bats attacked prey in the air and on the ground at similar rates, but success ratios 251
for aerial prey were more than twice of those of ground prey. Moreover, gleaning insects on 252
the ground most likely exposes bats to a higher predation risk from ground predators, and a 253
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
higher risk of injury (Toshkova et al. In prep; Brandmayr et al., 2009). This begs the question 254
of why most of the tagged bats (N = 22 of 34 bats) nonetheless preferred to capture insects on 255
the ground? To answer this we hypothesized that bats would choose the foraging strategy with 256
the highest profitability (i.e. energy intake/time) (Stephens and Krebs, 1986). We estimated 257
prey profitability by dividing estimated prey caloric values with prey search and handling time, 258
factoring in the success ratios for each strategy. We find that gleaning prey off the ground, 259
despite the much higher failure ratios, offers a prey profitability between 2.5 and 14 times 260
higher than aerial hawking owing to the much larger ground prey (Fig. 3F). Beetles gleaned 261
off the ground are 3-20 times heavier than aerial prey and have higher protein and fat 262
content(Razeng and Watson, 2015). In fact 85 % of the energy returns from a night of foraging 263
comes from gleaning despite the similar average number of prey capture attempts in air and on 264
the ground per night. Thus the paradox of preferring low success ratios and higher risk of injury 265
while gleaning (Fig. S10) is explained by a much higher prey profitability: the bats opt for a 266
high risk-high gain strategy. Nevertheless, we find that aerial hunting remains as a valuable 267
supportive foraging strategy for 19 out of 34 individuals on the nights sampled (Fig. 1DE), 268
contributing 15 % of the total intake for these bats. This reliance on aerial hawking next 269
prompted us to investigate how habitat affected prey profitability and the choice of foraging 270
strategy. 271
Prey profitability changes with habitat, but does not affect foraging decisions in greater 272
mouse-eared bats 273
It has been hypothesized that greater mouse-eared bats are opportunistic predators able to 274
maximize their energy intake by choosing the most profitable habitat (Arlettaz, 1999). We 275
tested this hypothesis by investigating how prey habitat (i.e. open fields vs forests (Arlettaz, 276
1999), Fig. 2GH) or movement style (i.e. commuting or active searching for prey, Fig. S2) 277
affect the profitability of prey and thereby influence foraging decisions of the wild bats (N = 7 278
GPS-tagged bats). The bats primarily gleaned prey during dedicated foraging bouts (Fig. S2A), 279
whereas hawking was used equally during commute and in foraging bouts (Fig. S2B). This 280
indicates that aerial hawking is a flexible strategy that is efficient enough to use during 281
commuting. This perhaps offsets some of the energetic costs of the often lengthy commutes 282
made by greater mouse-eared bats to preferred foraging areas while also potentially providing 283
them with concurrent information about aerial prey profitability. 284
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
When actively gleaning for prey, the tagged bats were more successful and attacked more 285
prey per bout in open fields than in forest habitats(26 vs 15 attacks/bout), but performed more 286
foraging bouts in the forest (field: 8 foraging bouts vs forest: 15 bouts; 3 bouts were a mixture 287
between field and forest). Thus, the quality of the gleaning habitat (here defined as a habitat 288
enabling higher success ratios and attack rates for the same strategy) does not appear to affect 289
the decision of which habitat to target. This could have several explanations: bats may be 290
maximizing energy intake by switching to forest habitat after depletion of open field habitats, 291
or they could be balancing predation risk and/or conspecific competition which are both likely 292
to be higher in open fields. In contrast to gleaning, bats were equally successful when using 293
echolocation to capture insects above open fields and below the canopy in forests revealing 294
that these bats are well able to hunt in semi-cluttered spaces(Stidsholt et al., 2021). Such a 295
flexible strategy allows them to be efficient hunters across different habitats equipping them to 296
exploit diverse environments. 297
Although this bat species hunt independently, the dominant foraging strategies of bats 298
tagged on the same night were more similar than for bats tagged on different nights. This 299
indicates that temporally-varying environmental parameters such as rain, wind or mass 300
emergence of aerial insects, experienced by all bats in the area on a given night, influence the 301
most beneficial foraging strategy. For example, after rain, rustling sounds of walking 302
arthropods on leaves are more difficult for bats to detect potentially halving the detection 303
range(Goerlitz, Greif and Siemers, 2008). Reduced detection distance could not only lower 304
detection rates but also capture success ratios (i.e. less time to plan and execute a capture). The 305
time between gleaning prey attacks increased by an average of 35 seconds on the two nights 306
with the most aerial captures (22nd of July 2018 and 2019, GLMM; testing whether the time 307
between gleaning attacks was explained by nights with a majority of aerial captures, p<0.01) 308
and the bats switched more often between gleaning and aerial strategies on an attack-to-attack 309
basis (GLMM; testing if the number of switches per night was explained by the nights with a 310
majority of aerial captures, p<0.01). Longer periods between prey attacks and more strategy 311
switching indicate that gleaning prey on these nights was less profitable compared to aerial 312
prey, either due to poor conditions on the ground or a mass emergence of aerial prey. 313
Our results demonstrate that greater mouse eared bats often resort to aerial hawking 314
despite their preference and specialisation for ground gleaning. Although much smaller prey 315
are taken on-the-wing, the consistently high success ratios and prey captures rates of hawking 316
independent of habitat make this a reliable backup strategy. The same qualities may make 317
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
hawking a more widespread strategy than expected in other gleaning bat species. Such foraging 318
flexibility both adds to the energy intake of wild bats and also improves their resilience to 319
changing conditions. Thus, this strategy might have allowed a wide range of bat species to tap 320
into the unpredictable, ephemeral food resources in open spaces. This would have put an 321
evolutionary pressure on maintaining aerial hawking capabilities to secure an additional food 322
resource in fluctuating environments, and may also explain why hawking remains as a foraging 323
strategy in many bat species traditionally seen as gleaning and trawling specialists. 324
Conclusion 325
We show that greater mouse-eared bats, a gleaning specialist, catch relatively few prey 326
items per night on the ground at high failure ratios, but still achieve a high prey profitability by 327
targeting large, energy-rich prey. This shows that prey availability and size, weighted by 328
relative hunting success, are important drivers of foraging decisions in wild bats. We find that 329
the bats use gleaning as a primary foraging tactic in a high risk-high gain approach, and switch 330
to aerial hunting when environmental changes reduce the profitability of ground prey. We 331
conclude that prey switching matched to environmental dynamics plays a key role in covering 332
the energy intake even in a specialised predator. 333
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Material and Methods 335
Method details 336
All experiments were carried out under the licenses: 721/12.06.2017, 180/07.08.2018 and 337
795/17.05.2019 from MOEW-Sofia and RIOSV-Ruse. We tagged and recaptured 34 female, post-338
lactating greater mouse-eared bats with sound-and-movement tags from late July to mid-August in 339
the seasons 2017, 2018 and 2019. The bats were caught with a harp trap at Orlova Chuka cavfe, 340
close to Ruse, NE-Bulgaria, in the early mornings as they returned to the roost. The bats were 341
kept at the Siemers Bat Research Station in Tabachka to measure the forearm lengths, CM3 342
and body weights (Table S1). Bats weighing above 29 grams were tagged and released the 343
following night between 10-11 p.m. at a field 8 km from the roost (Decimal degrees: 344
43.622097, 25.864917) or in the colony. The tags were wrapped in balloons for protection and 345
glued to the fur on the back between the shoulders with skin bond latex glue (Ostobond). The 346
bats on average spend 2 to 14 days equipped with the tags until we recaptured the bats at the 347
cave or the tags detached from the bats and fell to the ground below the colony. Upon recapture, 348
the bats were weighed and checked for any sign of discomfort from the tagging before they 349
were released back to the colony. 350
Tags 351
We used two different tags for this study. Both tags recorded continuous data during one night 352
of foraging. The first tag (Tag A) recorded audio data with an ultrasonic Knowles microphone 353
(FG-3329) at a sampling rate of 187.5 kHz, with 16 bit resolution, a 10 kHz 1-pole analog high-354
pass filter and a clipping level of 121 dB re 20µPa pk. These tags also included triaxial 355
accelerometers sampling the movement of the bats at 1000 Hz with a clipping level of 8 g. All 356
accelerometer data were calibrated, converted into acceleration units (m/s2) and decimated to 357
100 Hz. The orientation of the bats were recorded with triaxial magnetometers using a sampling 358
rate of 50 Hz. These tags weighed from 3.5-3.9 g (including VHF for localisation and recapture 359
of the bats). The second tag (Tag B) recorded audio using a MEMS microphone sampling at 360
94 kHz and with 16 bit resolution. The movement of the bats were recorded with triaxial 361
accelerometers at 50 Hz sampling rate and with a clipping level of 8 g. These tags also included 362
GPS sensors that logged the position of the bats every 15 seconds. These tags weighed 3.9-4.2 363
g (including VHF). In total, we have data from 16 bats with Tag A and 18 bats with Tag B (of 364
which 5 are 50 % duty cycled) (Table S1). 365
Tagging effects 366
Both tags weighed 11-15 % of the bodyweight of the tagged bats. We addressed the effects of 367
tagging on the data and the bats by the following procedures: (i) We trained bats in a flight 368
room to capture mealworms and moths tethered on strings and glean beetles of different sizes 369
from either a bowl or a square meter of natural forest floor (Table S2). The bats caught aerial 370
prey with high success ratios (95 % for mealworms and 69 % for moths (N = 2 bats)) with tags 371
and without tags (75 % for mealworms (N = 4 bats)). The bats gleaned beetles with a success 372
rate of 39 % from the square of forest floor. The success rate increased to 85 % when they were 373
catching beetles from a bowl with no escape options for the prey (N = 3 bats). This indicates 374
that the low gleaning success ratios in the wild are most likely not caused by tagging effects, 375
but more likely because beetles can escape in cracks or below leaves. Additionally, we could 376
not detect any visual difficulties with capturing prey from strings or beetles from the ground 377
(Video S1-2). (ii) The wild bats caught prey with high success ratios in air indicating that the 378
tag effects had little or no impact on the ability to capture prey. (iii) A previous study with the 379
same species and tags found that tagged and untagged bats spend equal amount of time foraging 380
(Egert-Berg, Hurme, Greif, Wilkinson, et al., 2018). (iv) The weight loss of these bats of 3-4 381
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
% during the instrumentation time was equal to the weight loss of control bats carrying only 382
VHF radio transmitters (0.5 g) indicating that handling and carrying of tags might disturb the 383
bats, but the additional extra load did not seem to add further energetic consequences to the 384
bats in addition to the VHF (Egert-Berg, Hurme, Greif, Goldstein, et al., 2018). 385
Quantification and statistical analysis 387
To understand how the tagged bats allocated their time and captured prey in the wild, all wild 388
tag recordings were manually analysed by displaying the acoustic and the movement data in 7-389
20 second segments with an additional option of playing back audio data. The visualisation 390
included three separate windows with synchronized data: (i) An envelope of audio data filtered 391
by a 20 kHz 4 pole high pass filter to detect the echolocation calls. (ii) A spectrogram of audio 392
data filtered by a 1 kHz 1 pole high pass filter to visualise the full bandwidth acoustic scene 393
showing echolocation calls, conspecific calls, chewing sounds, wind noise etc. (iii) The final 394
window showed triaxial accelerometer aiding the identification of wingbeats, landing, take-395
offs as well as capture events. 396
Categorisation of capture attempts 398
Greater mouse-eared bats are known to glean prey off the ground and to capture aerial prey. 400
To recognize gleaning capture attempts in the wild data, the tag recordings were ground truthed 401
by analysing sound and movement data from capture attempts of bats under controlled 402
experimental settings in the lab. Two individuals were trained to catch walking beetles on 403
vegetation using passive hearing while carrying a tag (Video S1). These ground captures were 404
identified by stereotyped patterns consisting of three simultaneous events (i) Low vocalizations 405
(around 50 dB re 20µPa2s) prior to the capture attempt indicating that the bat was using passive 406
listening to listen for prey generated cues. (ii) A short, broadband and loud audio transient 407
simultaneous to a peak in the accelerometer data indicating that the bat was landing on the 408
ground. (iii) The accelerometer signal indicating a landing was an increase in wingbeat 409
frequency and amplitude prior to the landing, a peak in the sway and heave axis (y and z 410
dimension, often with opposite values) at the time of contact with the ground and often 411
followed by a flattening of the signal on all three axes. These stereotyped audio and 412
accelerometer signals found in the laboratory experiments were matching signals seen in the 413
wild data (Figure S5). These signals were then identified in all wild tag recordings during the 414
visualisation and marking process. Aerial capture attempts were identified in the wild data if a 415
buzz was present. Only buzzes in flight were marked to exclude landing buzzes. In addition, 416
each capture attempt was marked as “successful” or “unsuccessful” based on the presence or 417
absence of chewing sounds. The chewing sounds were audible (Sound files are uploaded) and 418
for the low-noise tag recordings visible in spectrograms (Figure 3). For the five 50 % duty 419
cycled tag recordings we doubled the foraging attempts and successes. 420
Behavioural analysis: 421
To evaluate time allocation, we analysed data from 15 tags with both accelerometer and 422
magnetometer data. We separated times of rest from flight through identification of wingbeat 423
epochs. Wingbeats were detected as cyclic oscillations in the z-axis dimension of the 424
accelerometer data. We first band-pass filtered the z-axis dimension of accelerometer data 425
(from 5 to 25 Hz) by a delay-free symmetric FIR-filter (filter length: 1024 samples, sample 426
rate: 100 Hz). We then identified flight epochs as the time intervals where a running mean of 427
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
50 seconds of the wingbeat data were above a threshold of 20 m/s2. A window length of 50 428
seconds was chosen to avoid short, flight epochs consisting of only few wingbeats. 429
We identified times of foraging from travelling during flight epochs by changes in heading 430
because the bats fly straight towards and between foraging grounds (Egert-Berg, Hurme, Greif, 431
Goldstein, et al., 2018). Heading was computed by gimballing low-pass (3 Hz) filtered triaxial 432
magnetic field measurements with the pitch and roll estimated from down-sampled and low 433
pass filtered (3 Hz) accelerometer data (Johnson and Tyack, 2003). We applied a running mean 434
of 50 seconds to the heading measurements to evaluate foraging bouts. We chose a length of 435
50 seconds similar to the minimum foraging bout length used by Hurme et al., (2019) and in 436
our GPS analysis. Foraging bouts were identified as time intervals where the envelope of the 437
signal was above a threshold of 0.05. Due to the large variation between animals, this threshold 438
was raised to 0.3 for six tag recordings to avoid more than 10 switches between travelling and 439
foraging per night as used by Hurme et al., (2019) and in our GPS analysis. We omitted 440
identified foraging bouts with no capture attempts from the analysis (N = 62 out of 202 foraging 441
bouts for 16 bats). 442
GPS and habitat analysis 443
We recaptured 7 tags with GPS data. We used first-passage time analysis (Fauchald and 444
Tveraa, 2003) to identify foraging bouts from travelling bouts (Hurme et al., 2019) by using 445
the R package “adeHabitatLT” (Calenge, 2006). First, we converted the latitude and longitude 446
coordinates from degrees to meters (“proj4string” function in adeHabitatLT) and then 447
regularised the GPS tracks to 10 second time stamps. We then calculated the first-passage times 448
for all radii ranging from 5 to 400 m (in 5 m steps). We plotted the variance of the log-449
transformed first-passage times and found the highest value of around 250 m for all bats which 450
is in accordance with a previous study on Myotis vivesi (Hurme et al., 2019). This value 451
estimates the scale at which the bat is operating, and was used for all tracks. We used the 452
Lavielle method (“lavielle” in adeHabitatLT) to divide the path segments into foraging and 453
non-foraging bouts (Calenge, 2006). The minimum number of locations in a bout was chosen 454
to 5 (corresponding to 50 seconds), and the maximum number of segments per night was 50. 455
The function “chooseseg” was chosen to find the number of segments at which the contrast 456
between bouts were highest. The number of segments per night was estimated to either 10 or 457
11 per bat. To find a threshold value to separate foraging and travelling bouts, we found that 458
the distance travelled per segment showed bimodal distribution (function “bimodalitycoeff” in 459
Matlab (Zhivomirov, 2022)). We then fitted a Gaussian mixture model with two components 460
to the data (Figure S6). We defined the threshold between the two distributions as the lowest 461
quartile of the non-foraging (i.e. travelling) segments (Figure S6) corresponding to 40 meters 462
travelled per 10 second segment. 463
To determine in which habitats the bats were foraging, we transferred all tracks onto a Google 464
Map Satellite Imagine (using: “plot_google_map” in Matlab version 2021b). We manually 465
determined whether the GPS locations for each segment were located in one of three categories: 466
Field, Forest or both/others. We excluded foraging (N = 4) and non-foraging (N=23) bouts that 467
were not assigned to either field or forest. 468
Statistical analysis 469
All statistical modelling was performed in R (version 4.0.3). We fitted different models to 470
understand how capture attempts and success ratios were influenced by the foraging strategy, 471
the habitat and the mode of action (commuting vs foraging in bouts). For all models, we used 472
a goodness of fit evaluation based on the marginal (R(m)2) and the conditional R2 (R(c)2) 473
(Nakagawa and Schielzeth, 2013), the 0.05 criterion for statistical significance and Bat ID as 474
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
random effect. We examined potential collinearity between the predictor variables of each 475
model using variance inflation factor (from R-package “car” (Fox and Weisberg, 2019)). No 476
collinearity was found. 477
Model 1: We investigated how capture attempts and foraging successes were influenced by the 478
foraging strategy (N=29, the five 50 % duty cycled tags were omitted from the analysis). We 479
used foraging strategy as predictor variable and fitted two linear mixed effect model to the data 480
(“lme4” R-package (Bates et al., 2015)). In the first model (1a), we used foraging attempts as 481
response variable assuming a Poisson distribution (link=”log”), and in the second model (1b), 482
we used foraging success ratios as a normally distributed response variable. We examined the 483
residuals (“DHARMa” package in R) and found no deviations from the expected distribution. 484
Model 1a showed that the amount of capture attempts in a night was explained mostly by the 485
individual bat (random effect explained 70 % of the variance of the data) and less by the 486
foraging strategy (27 %) (Table S3). Overall, model 1b explained 77 % of the deviance of 487
foraging success. Subtracting the random effect of individual bat only decreased the explained 488
deviance by 8 % (Table S4). Model 1b revealed that bats are more successful when hawking 489
for prey. 490
Model 2: We investigated how the capture attempts and foraging successes were influenced by 491
the habitats of the seven bats tagged with GPS tags (Tag B). For both models, we used three 492
categorical predictor variables: habitat type (field vs forest), foraging strategy (aerial vs 493
hawking) and movement style (commuting vs foraging in bouts). 494
In model 2a, we used capture attempts (sum of failed and successful capture attempts in each 495
foraging bout) as a response variable and fitted a GLMM (“glmer” function in R-package 496
“lme4”) to the data with a Poisson distribution of the response variable and a link “sqrt” 497
function. In model 2b, we used foraging success (per foraging bout) as a normally distributed 498
response variable and fitted a LMM to the data (“lme” function in R-package “nlme” (Pinheiro 499
et al., 2022)). We used model selection procedures (“dredge” in R-package (Burnham and 500
Anderson, 2002)) to examine the best-fitted models using the AICc (corrected Akaike 501
information criterion). The best-fitted models included all three predictor variables in both 502
model 2a and 2b. We examined the residuals (“DHARMa” package in R) and slight deviations 503
from the expected distributions were found. 504
In model 2a, the three predictor variables only explained 15.0 % of the data, while the random 505
effects explained 78.1 % of the deviance in capture attempts. Thus, model 2a revealed that the 506
variance in capture attempts per night is largely explained by individual bat rather than habitat, 507
strategy or movement style (Table S5). In model 2b, the three predictor variables explained 508
72.3 % of the deviance of the success ratios. No difference was found when subtracting the 509
random effect for model 2b. This model revealed that habitat and strategy strongly influence 510
the success ratios of the bats (Table S6). 511
Model 3: We tested whether the immediate environment affected the foraging decision in bats 512
by investigating the relationship between the dominant foraging strategy and the night of 513
tagging for all 34 bats. The dominant foraging strategy was found as the ratio between aerial 514
and foraging attempts ranging from 0 to 1 (0 meaning 100 % aerial foraging; 1 meaning 100 515
% gleaning) and was used as a normally-distributed response variable. The data was fitted with 516
a LMM (function: “nlm” in R (Pinheiro et al., 2022)) using release night as a categorical fixed 517
effect, and release site as a categorical random effect. This model showed that there was strong 518
evidence of individual night on the chosen foraging strategy, indicating that bats on the same 519
nights choose the same foraging strategy. (N = 10 different nights, 34 bats, and a mean of 3 520
(2.1 SD) tags per night (Figure S2, Table S7)). 521
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Relative prey sizes based on quantification of mastication sounds 522
After the visualisation and marking of all capture attempts, a custom-written chewing detector 523
was used to automatically identify mastication sounds for nine tag recordings with sound 524
quality to perform this analysis. This analysis included 452 ground captures attempts and 387 525
aerial capture attempts. The detector was used in two steps: i) To automatically determine 526
whether the capture attempt was successful or not to verify the manual decision process based 527
on listening to the chewing sounds. ii) To characterise the mastication sounds of each prey item 528
from successful captures. 529
We extracted mastication sounds when the bat was flying after each capture attempt. Since the 530
bats in flight chewed between echolocation calls, we analysed the intercall interval from the 531
time of capture to either next capture attempt or 100 seconds ahead. We first detected 532
mastication sounds and then classified the mastication sounds for each intercall interval. 533
Each intercall interval was extracted, filtered by a 7 to 15 kHz 4-pole Butterworth filter and 534
convolved with a 40 ms Hanning window to exclude transients. These parameters were based 535
on mastication sounds from capture bats with peak frequencies of 7 kHz which corresponds to 536
the same peak frequency in wild bats. The intercall interval was classified as containing a 537
mastication sound if the maximum amplitudes of the filtered signals were above a threshold of 538
0.012 and the peak frequency was above 5 kHz and below 20 kHz (Figure S7-8 black vs red). 539
The intercall intervals that included mastication sounds were then used for the classification. 540
Here we filtered the intercall intervals of the original sound data containing mastication with a 541
5 kHz 4-pole high-pass filter to reduce flow noise. To extract the onset of chewing as well as 542
the duration, the detector also automatically extracted the time at which the bat produced the 543
10th and 90th quantile of the chewing sounds (Fig S7-8, grey dashed lines). The 10th and 90th 544
quantile were a conservative choice to avoid false detections. The onset of the chewing was 545
determined when the bat emitted the first sound in this interval. The duration of the chewing 546
was determined as the length of this interval. The handling time was estimated as sum of the 547
onset and duration of the mastication. 548
To test the performance of the classifier, the automatic and manual classification of successful 549
vs unsuccessful prey captures were compared. A confusion matrix was made separately for all 550
ground and aerial capture attempts. The classifier was evaluated by calculating the positive 551
predictive value (PPV) and the false negative rate (FNR) based on the derivatives of the 552
confusion matrix from the classification of the ground and aerial capture attempts (Table S8). 553
Overall, the detector worked with PPV values above 0.99. However, the FNR was higher for 554
aerial captures (3 % for ground; 9 % for aerial captures). The maximum sound energy in 555
chewing after aerial captures were weaker than after ground captures which may explain the 556
worse performance in air. To verify the mastication detector, we extracted and characterised 557
mastication sounds of known prey types and sizes in controlled setups in the laboratory where 558
the bats were eating while flying after prey captures (Figure S9). 559
DNA metabarcoding 560
To understand the taxonomic diversity of the diet of wild mouse-eared bats we collected faecal 561
samples from 54 bats (n= 26 in 2017, n= 28 in 2019, n = 48 female bats) returning to the roost 562
after foraging. Individual bats were caught with a harp-trap positioned at the entrance of Orlova 563
chucka cave, Pepelina, Bulgaria and placed in individual clean cotton bags until defecating. 564
Faecal samples were collected in 98% alcohol and stored until further analysis. DNA 565
extraction, data sequencing and bioinformatics were done following Morinière et al., 2016 (see 566
also Morinière et al., 2019). In short, DNA from the faecal samples were extracted by using 567
the DNEasy blood & tissue kit (Qiagen) following the manufacturer’s instructions. Multiplex 568
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
PCR was performed using 5 µL of extracted genomic DNA and high-throughput sequencing 569
(HTS)-adapted mini-barcode primers targeting the mitochondrial CO1 region. HTS was 570
performed on an Illumina MiSeq v2 (Illumina Inc., San Diego, USA) at AIM - Advanced 571
Identification Methods GmbH, Leipzig, Germany. Further, FASTQ files were combined and 572
sequence processing was performed with the VSEARCH v2.4.3 suite (Rognes et al., 2016) and 573
cutadapt v1.14 (Martin, 2011). Quality filtering was performed with the fastq_filter program 574
of VSEARCH, fastq_maxee 2; a minimum length of 100 bp was allowed. Sequences were 575
dereplicated with derep_fulllength, first at the sample level and then concatenated into one 576
fasta file, which was then dereplicated. Chimeric sequences were then detected and filtered out 577
from the resulting file. The remaining sequences were clustered into OTUs (Operational 578
Taxonomic Units) at 97% identity. To reduce likely false positives, a cleaning step was 579
employed that excluded read counts in the OTU table of less than 0.01% of the total read 580
number. OTUs were blasted against a custom Animalia database downloaded from BOLD 581
(Barcode Of Life Database, and BIN (Barcode Index Number) 582
information. 583
We measured the lengths of representatives of each species from online photo databases 584
( or the reference collection of The National Museum of Natural History 585
Sofia (Bulgaria) covering the same region in Bulgaria. The lengths used for the calculations 586
were estimated as the maximum values of the measured prey lengths. 587
Caloric value estimations: 589
To estimate the energetic intake of one night per bat, we first used length-weight regressions 590
to covert arthropod body-lengths to body masses (Straus and Avilés, 2018). Here we used the 591
weighted average of the two primary arthropod orders from each foraging strategy: ground 592
diet (i.e. Carabidae (78 %) and Orthoptera (22 %)) and aerial diet (i.e. Diptera (65 %) and 593
Lepidoptera (35%)) (Straus and Avilés, 2018). From this calculation, we estimated the dry 594
body masses of ground and aerial prey only based on metabarcoding data. In addition to this, 595
we also calculated the dry body mass of aerial prey based on the ratio between the 596
mastication (number of mastications/capture) after ground and aerial prey. This gave us an 597
additional value for the dry body mass of aerial prey. Thus, we proceeded with a higher and 598
lower estimate of aerial prey dry masses in the following calculations. 599
To convert dry body masses (mg/prey) into caloric values (J/prey), we multiplied the dry 600
body masses of the prey with the caloric values for either ground (Bell, 1990; Zygmunt, 601
Maryansky and Laskowski, 2006) or aerial (Kurta and Kunz, 1987; Bell, 1990) prey. Here, 602
we used the caloric values of the weighted average of the two most numerous prey orders for 603
ground and aerial prey. Nightly caloric intake of each bat was then calculated by multiplying 604
the number of eaten ground and aerial prey items with the caloric values of each prey type. 605
Profitability index calculations 606
To compare the profitability between the two foraging strategies, we first calculated the relative 607
profitability of each prey without including prey size using the following relationship: 608
()=   ()
( ()+   () )   (), 609
i = foraging strategy, Success rate (Mean success rate per bat (per bout with more than one 610
capture) as a fraction from 0 to 1), Time between prey attacks (mean number of seconds 611
between prey attacks per bat per night with unit seconds). Prey caloric values was estimated 612
using both metabarcoding and mastication analysis (see Caloric value estimations). The prey 613
profitability unit is therefore J/second per prey capture. 614
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Acknowledgements We are thankful to Kaloyana Kosseva for help with bat feces collection 615
in the field and to Ilias Foskolos and Nor Amira Abdul Rahman for assistance with laboratory 616
experiments. We are grateful to the entire crew at the Siemers Bat Research Station for the 617
support during the seasons 2017-2019 and to the Directorate of the Rusenski Lom Nature Park, 618
Bulgaria. 619
Author contributions L.S. was responsible for tagging data collection and analysis, 620
interpretation, and drafting of the manuscript. S.G. was responsible for conceptualization, 621
tagging data collection, and interpretation of the data. A.H. collected the DNA metabarcoding 622
data, analyzed and interpreted the data. H.R.G. was responsible for conceptualization and 623
interpretation of the data. M.J. designed and manufactured the tags and contributed to the 624
interpretation of the data. Y.Y. designed the tagging experiment and was responsible for 625
conceptualization. P.T.M. was responsible for conceptualization and interpretation of the data. 626
All authors contributed to the writing of the manuscript. 627
Declaration of interests The authors declare that they have no competing interests. This study 628
was funded by the Carlsberg Semper Ardens grant to P.T.M., by the Emmy Noether program 629
of the Deutsche Forschungsgemeinschaft (DFG; German Research Foundation, grant no. 630
241711556) to H.R.G, and by the Bulgarian Academy of Sciences to A.H (Grant No. DFNP-631
17-71/28.07.2017). All experiments were carried out under the following licenses: 632
721/12.06.2017, 180/07.08.2018, and 795/17.05.2019 issued from the Ministry of the 633
Environment and Waters, Bulgaria. 634
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
References 637
Aliperti, J. R. et al. (2017) ‘Using behavioral and stable isotope data to quantify rare dietary 638
plasticity in a temperate bat’, Journal of Mammalogy, 98(2), p. gyw196. doi: 639
10.1093/jmammal/gyw196. 640
Arlettaz, R. (1996) ‘Feeding behaviour and foraging strategy of free-living mouse-eared bats, 641
Myotis myotis and Myotis blythii’, Animal Behaviour, 51, pp. 1–11. 642
Arlettaz, R. (1999) ‘Habitat selection as a major resource partitioning mechanism between 643
the two sympatric sibling bat species Myotis myotis and Myotis blythii ’, Journal of Animal 644
Ecology, 68(3), pp. 460–471. doi: 10.1046/j.1365-2656.1999.00293.x. 645
Bates, D. et al. (2015) ‘Fitting Linear Mixed-Effects Models Using lme4’, Journal of 646
Statistical Software, 67(1), pp. 1–48. Available at: 10.18637/jss.v067.i01. 647
Bell, G. P. (1982) ‘Behavioral and ecological aspects of gleaning by a desert insectivorous 648
bat Antrozous pallidus (Chiroptera: Vespertilionidae)’, Behavioral Ecology and 649
Sociobiology, 10(3), pp. 217–223. doi: 10.1007/BF00299688. 650
Bell, G. P. (1990) ‘Birds and mammals on an insect diet: a primer on diet composition 651
analysis in relation to ecological energetics’, Studies in Avian Biology, 13, pp. 416–422. 652
Brandmayr, P. et al. (2009) ‘The evolution of defence mechanisms in carabid beetles: a 653
review’, in Casellato, S., Burighel, P., and Minelli, A. (eds) Life and Time: The Evolution of 654
Life and its History. Padova: CLEUP, pp. 25–43. 655
Burnham, K. P. and Anderson, D. R. (2002) Model selection and multimodel inference: a 656
practical information-theoretic approach. 2nd edn. New York: Springer-Verlag. 657
Calenge, C. (2006) ‘The package “adehabitat” for the R software: A tool for the analysis of 658
space and habitat use by animals’, Ecological Modelling, 197(3–4), pp. 516–519. doi: 659
10.1016/j.ecolmodel.2006.03.017. 660
Cvikel, N. et al. (2015) ‘Bats aggregate to improve prey search but might be impaired when 661
their density becomes too high’, Current Biology, 25(2), pp. 206–211. doi: 662
10.1016/j.cub.2014.11.010. 663
Egert-Berg, K., Hurme, E. R., Greif, S., Goldstein, A., et al. (2018) ‘Resource ephemerality 664
drives social foraging in bats’, Current Biology, 28(22), pp. 3667-3673.e5. doi: 665
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
10.1016/j.cub.2018.09.064. 666
Fauchald, P. and Tveraa, T. (2003) ‘Using first-passage time in the analysis of area-restricted 667
search and habitat selection’, Ecology, 84(2), pp. 282–288. doi: 10.1890/0012-668
9658(2003)084[0282:UFPTIT]2.0.CO;2. 669
Fenton, M. B. (1990) ‘The foraging ecology of animal-eating bats’, Canadian Journal of 670
Zoology, 68(3), pp. 411–412. 671
Fox, J. and Weisberg, S. (2019) An R Companion to Applied Regression: Appendices. Third, 672
Sage publications. Third. Thousand Oaks, CA: SAGE Publications Ltd. 673
Goerlitz, H. R., Greif, S. and Siemers, B. M. (2008) ‘Cues for acoustic detection of prey: 674
Insect rustling sounds and the influence of walking substrate’, Journal of Experimental 675
Biology, 211(17), pp. 2799–2806. doi: 10.1242/jeb.019596. 676
Hackett, T. D., Korine, C. and Holderied, M. W. (2014) ‘A whispering bat that screams: 677
Bimodal switch of foraging guild from gleaning to aerial hawking in the desert long-eared 678
bat’, Journal of Experimental Biology, 217(17), pp. 3028–3032. doi: 10.1242/jeb.100362. 679
Herrera, M. L. G. et al. (2001) ‘Sources of protein in two species of phytophagous bats in a 680
seasonal dry forest: Evidence from stable-isotope analysis’, Journal of Mammalogy, 82(2), 681
pp. 352–361. doi: 10.1644/1545-1542(2001)082<0352:SOPITS>2.0.CO;2. 682
Hurme, E. R. et al. (2019) ‘Acoustic evaluation of behavioral states predicted from GPS 683
tracking: a case study of a marine fishing bat’, Movement Ecology, 7(21), pp. 1–14. 684
Johnson, M. P. and Tyack, P. L. (2003) ‘A digital acoustic recording tag for measuring the 685
response of wild marine mammals to sound’, IEEE Journal of Oceanic Engineering, 28(1), 686
pp. 3–12. doi: 10.1109/JOE.2002.808212. 687
Kleiber, M. (1947) ‘Body size and metabolic rate’, Physiological reviews, 27(4), pp. 511–688
541. 689
Kline, E., Ripperger, S. P. and Carter, G. G. (2021) ‘Habituation of common vampire bats to 690
biologgers’, Royal Society Open Science, 8(211249), pp. 1–6. Available at: 691 692
Kolb, A. (1958) ‘Nahrung und nahrungsaufnahme bei fledermäusen’, Z. Säugetierk, 23, pp. 693
84–95. 694
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Kunz, T. H. et al. (2011) ‘Ecosystem services provided by bats’, Annals of the New York 695
Academy of Sciences, 1223(1), pp. 1–38. doi: 10.1111/j.1749-6632.2011.06004.x. 696
Kurta, A. et al. (1989) ‘Energetics of pregnancy and lactation in free-ranging little brown bats 697
(Myotis lucifugus)’, Physiological Zoology, 62(3), pp. 804–818. doi: 698
10.1086/physzool.62.3.30157928. 699
Kurta, A. and Kunz, T. H. (1987) ‘Size of bats at birth and maternal investment during 700
pregnancy.’, Symposia of the Zoological Society of London, 57, pp. 79–106. Available at: 701 (Accessed: 17 June 2018). 702
Norberg, U. M. and Rayner, J. M. V. (1987) ‘Ecological morphology and flight in bats 703
(Mammalia; Chiroptera): wing adaptations, flight performance, foraging strategy and 704
echoloction’, Philosophical Transactions of The Royal Society, 316(1179), pp. 335–427. 705
Pinheiro, J. et al. (2022) ‘nlme: Linear and nonlinear mixed effects models.’, R package 706
version, 3.1-158. Available at: 707
Portugal, S. J. and White, C. R. (2018) ‘Miniaturization of biologgers is not alleviating the 708
5% rule’, Methods in Ecology and Evolution, 9(7), pp. 1662–1666. doi: 10.1111/2041-709
210X.13013. 710
Ratcliffe, J. M. and Dawson, J. W. (2003) ‘Behavioural flexibility: The little brown bat, 711
Myotis lucifugus, and the northern long-eared bat, M. septentrionalis, both glean and hawk 712
prey’, Animal Behaviour, 66(5), pp. 847–856. doi: 10.1006/anbe.2003.2297. 713
Ratcliffe, J. M., Fenton, M. B. and Shettleworth, S. J. (2006) ‘Behavioral flexibility 714
positively correlated with relative brain volume in predatory bats’, Brain, Behavior and 715
Evolution, 67(3), pp. 165–176. doi: 10.1159/000090980. 716
Razeng, E. and Watson, D. M. (2015) ‘Nutritional composition of the preferred prey of 717
insectivorous birds: Popularity reflects quality’, Journal of Avian Biology, 46(1), pp. 89–96. 718
doi: 10.1111/jav.00475. 719
Rydell, J., McNeill, D. P. and Eklöf, J. (2002) ‘Capture success of little brown bats (Myotis 720
lucifugus) feeding on mosquitoes’, Journal of Zoology, 256(3), p. S0952836902000419. doi: 721
10.1017/S0952836902000419. 722
Schnitzler, H.-U. and Kalko, E. K. V. (2001) ‘Echolocation by insect-eating bats’, 723
BioScience, 51(7), p. 557. doi: 10.1641/0006-3568(2001)051[0557:EBIEB]2.0.CO;2. 724
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
Simmons, N. B. (2005) ‘An Eocene big bang for bats’, Science, 307(5709), pp. 527–528. doi: 725
10.1126/science.1108871. 726
Stephens, D. and Krebs, J. R. (1986) Foraging Theory. Edited by John R. Krebs and T. H. 727
Clutton-Brock. Princeton, New Jersey: Princeton University Press. 728
Stidsholt, L. et al. (2018) ‘A 2.6-g sound and movement tag for studying the acoustic scene 729
and kinematics of echolocating bats’, Methods in Ecology and Evolution, (October), pp. 1–730
11. doi: 10.1111/2041-210X.13108. 731
Stidsholt, L. et al. (2021) ‘Hunting bats adjust their echolocation to receive weak prey echoes 732
for clutter reduction’, Science Advances, 7(eabf1367). doi: 10.1126/sciadv.abf1367. 733
Straus, S. and Avilés, L. (2018) ‘Estimating consumable biomass from body length and order 734
in insects and spiders’, Ecological Entomology, 43(1), pp. 69–75. doi: 10.1111/een.12471. 735
Teeling, E. C. et al. (2005) ‘A Molecular Phylogeny for Bats Illuminates Biogeography and 736
the Fossil Record’, Science, 307, pp. 580–584. 737
Zhivomirov, H. (2022) ‘Bimodality Coefficient Calculation with Matlab, MATLAB Central 738
File Exchange. Retrieved February 8.’ 739
Zygmunt, P. M. S., Maryansky, M. and Laskowski, R. (2006) ‘Body mass and caloric value 740
of the ground beetle (Pterostichius oblongopunctatus) (Coleoptera, Carabidae) along a 741
gradient of heavy metal pollution’, Environmental Toxicology and Chemistry, 25(10), pp. 742
2709–2714. 743
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for thisthis version posted October 26, 2022. ; bioRxiv preprint
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Rapid advancements in biologging technology have led to unprecedented insights into animal behaviour, but testing the effects of biologgers on tagged animals is necessary for both scientific and ethical reasons. Here, we measured how quickly 13 wild-caught and captively isolated common vampire bats ( Desmodus rotundus ) habituated to mock proximity sensors glued to their dorsal fur. To assess habituation, we scored video-recorded behaviours every minute from 18.00 to 06.00 for 3 days, then compared the rates of grooming directed to the sensor tag versus to their own body. During the first hour, the mean tag-grooming rate declined dramatically from 53% of sampled time (95% CI = 36–65%, n = 6) to 16% (8–24%, n = 9), and down to 4% by hour 5 (1–6%, n = 13), while grooming of the bat's own body did not decline. When tags are firmly attached, isolated individual vampire bats mostly habituate within an hour of tag attachment. In two cases, however, tags became loose before falling off causing the bats to dishabituate. For tags glued to fur, behavioural data are likely to be impacted immediately after the tag is attached and when it is loose before it falls off.
Full-text available
How animals extract information from their surroundings to guide motor patterns is central to their survival. Here, we use echo-recording tags to show how wild hunting bats adjust their sensory strategies to their prey and natural environment. When searching, bats maximize the chances of detecting small prey by using large sensory volumes. During prey pursuit, they trade spatial for temporal information by reducing sensory volumes while increasing update rate and redundancy of their sensory scenes. These adjustments lead to very weak prey echoes that bats protect from interference by segregating prey sensory streams from the background using a combination of fast-acting sensory and motor strategies. Counterintuitively, these weak sensory scenes allow bats to be efficient hunters close to background clutter broadening the niches available to hunt for insects.
Full-text available
Background Multiple methods have been developed to infer behavioral states from animal movement data, but rarely has their accuracy been assessed from independent evidence, especially for location data sampled with high temporal resolution. Here we evaluate the performance of behavioral segmentation methods using acoustic recordings that monitor prey capture attempts. Methods We recorded GPS locations and ultrasonic audio during the foraging trips of 11 Mexican fish-eating bats, Myotis vivesi, using miniature bio-loggers. We then applied five different segmentation algorithms (k-means clustering, expectation-maximization and binary clustering, first-passage time, hidden Markov models, and correlated velocity change point analysis) to infer two behavioral states, foraging and commuting, from the GPS data. To evaluate the inference, we independently identified characteristic patterns of biosonar calls (“feeding buzzes”) that occur during foraging in the audio recordings. We then compared segmentation methods on how well they correctly identified the two behaviors and if their estimates of foraging movement parameters matched those for locations with buzzes. Results While the five methods differed in the median percentage of buzzes occurring during predicted foraging events, or true positive rate (44–75%), a two-state hidden Markov model had the highest median balanced accuracy (67%). Hidden Markov models and first-passage time predicted foraging flight speeds and turn angles similar to those measured at locations with feeding buzzes and did not differ in the number or duration of predicted foraging events. Conclusion The hidden Markov model method performed best at identifying fish-eating bat foraging segments; however, first-passage time was not significantly different and gave similar parameter estimates. This is the first attempt to evaluate segmentation methodologies in echolocating bats and provides an evaluation framework that can be used on other species.
Full-text available
Current models used to estimate insect prey biomass for diet studies use whole weight. However, a large proportion of an arthropod's body is taken up by an indigestible exoskeleton, leading to erroneous estimation of the food intake of insectivorous animals. 2. Linear mixed effect models were used to obtain equations to predict consumable biomass from body length for a variety of Neotropical insects and spiders. These data were obtained by feeding taxa of various orders to groups of 100 social spiders and comparing pre- and post-consumption weights using size-matched controls. 3. Significant linear relationships were found relating body size to consumed biomass for all orders, with slopes ranging from 1.276 to 4.011 and R2 values from 0.476 to 0.929. For orders other than spiders and Orthoptera, the increase in weight with size exhibited negative allometric scaling, suggesting a decrease in tissue density, or an increase in internal air space, with size. 4. Although there were significant differences across taxonomic orders in the proportion of biomass consumed, within most orders the proportion consumed did not differ significantly with body size. The estimated regression coefficients may be used by other workers to estimate consumable biomass of arthropod prey for studies requiring large sample sizes or non-lethal sampling of rare or endangered species.
Full-text available
Maximum likelihood or restricted maximum likelihood (REML) estimates of the parameters in linear mixed-effects models can be determined using the lmer function in the lme4 package for R. As for most model-fitting functions in R, the model is described in an lmer call by a formula, in this case including both fixed- and random-effects terms. The formula and data together determine a numerical representation of the model from which the profiled deviance or the profiled REML criterion can be evaluated as a function of some of the model parameters. The appropriate criterion is optimized, using one of the constrained optimization functions in R, to provide the parameter estimates. We describe the structure of the model, the steps in evaluating the profiled deviance or REML criterion, and the structure of classes or types that represents such a model. Sufficient detail is included to allow specialization of these structures by users who wish to write functions to fit specialized linear mixed models, such as models incorporating pedigrees or smoothing splines, that are not easily expressible in the formula language used by lmer.
1.To study sensorimotor behaviour in wild animals, it is necessary to synchronously record the sensory inputs available to the animal, and its movements. To do this, we have developed a biologging device that can record the primary sensory information and the associated movements during foraging and navigating in echolocating bats. 2.This 2.6 ‐gram tag records the sonar calls and echoes from an ultrasonic microphone, while simultaneously sampling fine‐scale movement in three dimensions from wideband accelerometers and magnetometers. In this study, we tested the tag on an European noctula (Nyctalus noctula) during target approaches and on four big brown bats (Eptesicus fuscus) during prey interception in a flight room. 3.We show that the tag records both the outgoing calls and echoes returning from objects at biologically relevant distances. Inertial sensor data enables the detection of behavioural events such as flying, turning, and resting. In addition, individual wing‐beats can be tracked and synchronized to the bat's sound emissions to study the coordination of different motor events. 4.By recording the primary acoustic flow of bats concomitant with associated behaviours on a very fine time‐scale, this type of biologging method will foster a deeper understanding of how sensory inputs guide feeding behaviours in the wild. This article is protected by copyright. All rights reserved.
This chapter gives results from some illustrative exploration of the performance of information-theoretic criteria for model selection and methods to quantify precision when there is model selection uncertainty. The methods given in Chapter 4 are illustrated and additional insights are provided based on simulation and real data. Section 5.2 utilizes a chain binomial survival model for some Monte Carlo evaluation of unconditional sampling variance estimation, confidence intervals, and model averaging. For this simulation the generating process is known and can be of relatively high dimension. The generating model and the models used for data analysis in this chain binomial simulation are easy to understand and have no nuisance parameters. We give some comparisons of AIC versus BIC selection and use achieved confidence interval coverage as an integrating metric to judge the success of various approaches to inference.
Food availability is emerging as a key determinant of avian occurrence and habitat use in a variety of systems, but insectivores have received less attention than other groups and the potential influence of nutritional quality has rarely been considered. Rather than a uniform food source, arthropods vary greatly in terms of nutritional composition, but does this variation translate into differential consumption? Building on previous work that demonstrated clear preference for some arthropod groups by 13 species of ground-foraging insectivores, we compare the nutritional composition of these arthropod groups with other groups commonly encountered but seldom consumed in the same habitat types. Using samples of arthropods collected from a eucalypt woodland in southern Australia, we found the high frequency prey groups (Coleoptera, Lepidoptera, Orthoptera and Araneae) consistently contained higher fractions of crude protein and total fat than the low frequency groups (Diptera, Hymenoptera and Odonata). Even more clear-cut differences were noted in terms of micronutrients; high frequency prey containing significantly greater concentrations of seven elements than low frequency prey and significantly greater amounts per individual arthropod for all eleven elements measured. These results indicate that the nutritional quality plays an important role in prey selection in insectivores and suggests that micronutrients may be more important determinants of prey choice than previously recognized. Integrating these findings with previous work suggesting food limitation may constrain distribution patterns of birds in fragmented landscapes, we contend that variation in nutritional quality helps explain observed patterns in insectivore diets and occurrence. In addition to explaining why smaller and more disturbed habitats are unable to support resident insectivore populations, this bottom-up mechanism may underlie the disproportionate sensitivity of insectivores to land-use intensification.