Discussion
Bat point counts versus traditional sampling
methods
For sufficient sampling effort (i.e. >2.5 sampling hours or
>20 individuals), the bat point counts had a higher
probability to detect new species than traditional methods, potentially
better representing the total species pool. In our study, mist-netting
did not perform well due to the hyper-dominance of Cynopterus
brachyotis in mist net captures, and passive acoustic monitoring cannot
detect the five non-echolocating bat species. Logistical constraints did
not allow for longer sampling durations in bat point counts to detect
more species and increase confidence of our estimates, and additional
personnel could have introduced a sampling bias. Nonetheless, in one
quarter of the sampling duration of the other sampling methods, bat
point counts were tied in the first place with mist-netting for raw taxa
counts. Given that the bat point counts sample was almost complete in
the shortest time and allowed to detect relatively more species than
using the other methods, the new bat point count method can be
considered the most effective and time-efficient. It follows that bat
point counts could be especially attractive for rapid assessments or for
researchers with time constraints in the field.
Theoretically, passive collection of complementary acoustic and visual
data enables bat point counts to detect echolocating and
non-echolocating bat types without bias. Correspondingly, point counts
could be particularly useful as a single sampling method in the
Paleotropics and Oceania, where both types coexist. As a consequence, in
our study, they reached the highest species richness at equal sampling
coverage with the traditional sampling methods. Admittedly, if the aim
was to detect a maximum number of species, combining mist nets with
passive recording could yield better results than bat point counts.
However, the logistical effort and expertise requirements should not be
underestimated, and bat incidences from studies based on these different
methods are not directly comparable. As a result, bat point counts are
currently the only way to obtain unbiased prevalences of echolocating
and non-echolocating bats, foregoing potential methodological and
taxonomic incompatibilities between sampling methods.
By design, bat point counts should have a smaller taxonomic sampling
bias compared to established sampling methods. All bats must be
detectable thermally as they are metabolically active, hence warmer than
the surrounding environment. Their detectability depends mainly on their
size and distance to the thermal scope. The larger bats we caught were
approximately 10 cm large (from head to tail - roughly the palm of the
hand used to determine detection distances), and could be detected -
albeit presumably not identified - at up to 80 m. Geometrically, 4 cm
large bats would thus have a detection area of 32 m radius. In contrast,
bat diversity studies that are based on trapping and acoustics rarely
account for the detectability of different species when comparing
community across space. Ultrasound detection ranges are highly variable
and species-specific as they depend strongly on the frequency, sound
level, and directivity of the calls
(K. Darras et al.,
2016). Also mist-netting species abundances depend on their exact
setup, which cannot systematically be reproduced across studies, and
given the taxonomic sampling biases mentioned earlier, they are unlikely
to be comparable across species. It follows that with our approach,
specific thermal detection ranges are relatively easily measured and
likely have less biased detectability between species, so that the
corresponding density estimates can be computed more accurately and
compared across species.
Bat point counts revealed that in our region, there is a strong sampling
bias against insectivorous bats. Acoustic monitoring alone cannot reveal
the magnitude of that bias, and although mist netting potentially could,
four echolocating species were not caught with our mist nets at all -
even though Hipposideridae and Rhinolophidae were caught in other
studies (Fukuda et
al., 2009; Huang et al., 2014). Previous studies from oil palm
plantations in Southeast Asia used only mist nets so far, and perhaps as
a consequence, it is recurrently stated that they are dominated by
frugivorous bats - especially Cynopterus brachyotis(Azhar et al., 2015;
Fukuda et al., 2009; Mohd-Azlan, 2019; Syafiq et al., 2016). In
contrast, in bat point counts, only seven out of 100 detections came
from Pteropodidae, and this ratio might even be lower when considering
their higher detectability due to their larger size. These results are
consistent with the fact that oil palm monocultures do not provide food
for Pteropodidae; they usually fly through to forage on fig trees on
river banks (pers. obs. KD, EY). Thus, it appears that much of the bat
assemblage is ignored when using only mist nets, underlining their
strong taxonomic bias in our system
(Huang et al., 2019),
and suggesting similar biases in other tropical studies might exist. As
a consequence, the contribution of bat communities to regulating insect
populations in oil palm may also have been underestimated.
The three methods we tested have different practical requirements (Table
1). Taxonomic expertise is needed for all methods, but in the case of
point counts and automated ultrasound recordings, it can be delayed or
out-sourced: species identification can be done by experts using
computers, anytime. Data processing is time-effective for mist-netting
as data can typically be entered after the survey. However, sound
recordings must usually be retrieved, uploaded and annotated recordings,
and for bat point counts, photographs must additionally be processed
analogously. In our study with our iteratively developed workflow, we
estimate a post-survey workload that is approximately equal to the
duration of the acoustic recordings to annotate and identify bat passes.
For bat points counts, another two to five minutes processing time per
detection can be added to obtain detection statistics and identification
features, however, the acoustic processing can be considerably shortened
by focusing on thermal detections, which made up only 6 % of the total
recording time. Regarding initial costs, mist netting requires
considerable training, and point counts currently require high expertise
and expenses for the assembly of the sampling rig. In our study,
high-end hardware was already available, but we estimated lower costs
for all methods with alternative hardware at equal performance (Table
1): much cheaper, equivalent sound recorders can be obtained
(Hill et al., 2019),
and currently, a complete rig can be built for approximately 2200 EUR.
Although the initial hardware costs of bat point counts can be
prohibitive for less well-funded research projects, we strive to develop
the bat point counts rig further
(K. Darras et al.,
2021) to lower the costs and increase availability. In comparison to
mist-netting, sound recorder installation and bat point counts can be
carried out more easily by trained personnel, even alone, if safety at
night is no concern.
Application challenges and research
opportunities
Point counts potentially cover large sampling areas, but they depend on
the surveyed site, as clear lines of sight are required. In our study,
the sparse understory allowed us to detect bats at relatively long
ranges (48 m) that were only limited by larger obstacles such as palms
or uneven terrain. We can predict that in even, open terrain without
trees, the detection spaces would be even greater (approximately 80 m -
the maximum range we measured). However, in previous trials with a
prototype sampling rig in a forest with dense understory vegetation, the
detection range was more limited (approximately 18 m, pers. obs. EY).
Hence, we suggest choosing good vantage points or clearing lower, nearby
vegetation that considerably obstructs the field of view. More
importantly, irrespective of the variability of detection ranges,
detection ranges are ultimately measurable so that detectability
variations can be accounted for to obtain rough density estimates, an
approach that is still biased for mist netting studies, and rare for
acoustic studies. Theoretically, bat point counts could even be used to
derive more reliable density estimates from distance sampling
approaches: using the first detection of each species (a standard
approach for avoiding double-counting) and its angular size, detection
distances could be estimated. Finally, to solve the issue that only a
part of the surroundings are thermally sampled at any point in time,
thermal scopes with higher resolution and larger field of view could be
used to cover a larger detection area.
Bat point counts are a fundamentally different method that requires
human presence but does not capture any live specimens. Instead, users
must become proficient with the handling of acoustic and photographic
data. Our technical information and workflows
(K. Darras et al.,
2021) as well as our ecoacoustic software tools
(K. Darras et al.,
2020) and our identification key presented here facilitate that
process. However, different sampling regions - and to some degree,
habitats - currently require dedicated identification keys. So far,
identification keys are based on absolute external, and sometimes
internal morphological measurements and features, and none use acoustic
data to out knowledge. Researchers often cannot make complementary
ultrasound recordings when capturing bats, and flying-tent or
hand-release recordings can yield calls that are atypical of free-flying
bats (Dietz & Kiefer,
2015). However, we showed that it is precisely the complementarity of
acoustic and photographic data that improves taxonomic identification,
and we are the first to develop such an identification key that combines
acoustic and morphological features. We suggest that this becomes a
research priority.
Bat point counts will resolve more species and individuals as the
technology matures and yields better near-infrared imagery
(K. Darras et al.,
2021). One of our sonotypes remains unidentified, but it is likely that
we will be able to discriminate between its candidate species. Some
detections included in that sonotype likely belonged toMiniopterus sp., but this was based on the personal experience of
co-author JCH with that genus’ phenotype, which is challenging to
reproduce. At the moment, frugivores are lumped into theCynopterus genus because species identification requires internal
and absolute body metrics. Still, it is possible to distinguish them
from genera which have distinct head shapes and body sizes. Better
near-infrared imagery will increase the proportion of usable pictures,
and reveal the wing bones and face more clearly. This would further
improve the shape of the species accumulation curves and increase their
confidence. We also photographed detailed morphological features that
could aid in discriminating between individuals: we sexed a maleCynopterus by its visible penis, and indentified it as an
individual with a hole in its wing (Text S2). These individual
signatures would yield more realistic abundance estimates and provide
information for the conservation of wild populations.
Morphological and acoustic data from point counts could be invaluable
for resolving species complexes (consisting of several candidate
identifications). Analogously to our present attempts, we are hopeful
that sonotypes can be resolved with more taxonomic depth, such as for
“whispering” bat species (e.g., Phyllostomidae)
(Yoh et al., 2020).
The recordings obtained from bat point counts, using an ultrasonic horn,
were also more amenable to call libraries: 1) their signal-to-noise
ratio is consistently higher, as the horn amplifies the sounds from an
actively tracked bat; 2) call durations are more accurately measurable,
as echoes from the surroundings are shielded by the horn; 3) calls are
representative of free-flying bats, unlike calls from handheld bats,
bats flying in tents, and likely also calls from distraught, released
bats. It follows that bat point counts can yield reference data to
identify unresolved sonotypes inside unattended recordings, even a
posteriori. Conversely, acoustic data can resolve ”cryptic” species
complexes that are morphologically almost indistinguishable: although we
found only one member of the Hipposideros bicolor complex,Hipposideros kunzi was readily identified using its calls’
maximum energy frequency
(Murray et al.,
2018). Interestingly, mist-net captures could be used to generate
reference near-infrared reference photographs for facilitating
identification from pictures of free-flying bats, using coloration
differences that we were not able to use here. Potentially, the flight
pattern observed in the thermal scope can also be diagnostic
(K. F. A. Darras,
2020), but their usefulness must be evaluated more throughly.
The combination of direct visual observation with ultrasonic recording
allows to study bat behavior (Text S2). Interactions between individuals
(within and between species) can be observed - we saw several encounters
between bats. In some cases, the bats would fly together, and in other
cases, they would avoid each other. Possibly, the function of social
calls could be elucidated and linked to competition for critical
resources (Corcoran &
Conner, 2014) or partners
(Voigt et al., 2008),
or calls from individual bats with different ages, sexes, and group
memberships (Kao et
al., 2020; Pfalzer & Kusch, 2003; Siemers et al., 2005). Second,
flight maneuvers such as diving can be seen (Text S2), giving insights
about hunting behavior: we observed several dives and potentially a
catch on the wing. Also, the head position was variable, appearing to
indicate the echolocating direction for scanning prey and obstacles, and
helping their identification. Finally, the exact coupling of
photographic and audio data reveals what calls are emitted in which
situation or environment - for instance during a diving maneuver (Text
S3) - and when exactly bats emit feeding buzzes and social calls
(Middleton et al.,
2014). Previously, such observations were only possible in carefully
controlled artificial environments such as tunnels with extensive setups
(Clark, 2021).
Conclusion
Bat point counts are a new tool for ecologists and a promising avenue
for sampling flying bat communities comprehensively and efficiently.
Technological advances will lower the cost and increase the
practicability and efficacy of bat point counts in the near future
(K. Darras et al.,
2021). The method needs to be evaluated further in different
environments with more speciose bat assemblages. Still, mist-netting is
obviously needed for capturing and measuring undescribed species, for
taking samples, and for assessing the physiology; it currently delivers
the highest, usually species-level identification accuracy. Also
automated ultrasound recordings are an effective, standardised, and
practical way of sampling echolocating bats. Yet, bat point counts are
unique and they could potentially be used in conjunction with bird point
counts to comprehensively sample all flying vertebrates. The potential
of newer technologies should be embraced to advance chiropterology and
advance fundamental and applied research questions in ecology and
conservation. Bat point counts, as a direct observation method that
makes bats audible and visible, shines a new light on flying bat
communities and their behavior, and will potentially lead to radically
new insights.