Introduction

Biodiversity sampling is biased towards species that are easily and directly detectable with our human senses (Moussy et al., 2021). Even though remote visible light imagery has been used for decades (Blackwell et al., 2006; Cutler & Swann, 1999), newer technologies based on the detection of the broader electromagnetic energy spectrum are becoming more accessible and further facilitate detecting and identifying animals passively and remotely (Turner et al., 2003; Vance et al., 2016). The applications in ecology and biodiversity conservation have great potential for scientists and conservationists (Pimm et al., 2015), especially when sampling elusive animals. Here, we focus on the detection of bats (Chiroptera), a taxon that is notoriously difficult to sample because they are nocturnal, fast, and silent fliers. This partly explains the relative lack of knowledge about bats, although they are the second most diverse order of mammals, provide important, wide-ranging ecosystem services, and experience acute threats (Frick et al., 2020; Kunz et al., 2011).
Bats are typically studied by capture using traps or by roost surveys. Mist netting and harp-trapping are the most common sampling methods for bats outside of their roosts. They are valuable for measuring the bats’ morphology precisely, taking physical samples (blood, tissue, parasites), assessing their physiological status, and estimating bat abundance directly. However, they are logistically challenging and have biases: species flying above nets (e.g., large fruit bats) are rarely caught, nets are avoided by some echolocating bats (e.g. “whispering bats”), and other bats can learn to avoid them, requiring daily net moving (Marques et al., 2013). Harp traps are more effective for some species, but they have variable performance (Berry et al., 2004) and may be more useful in South-east Asia (Furey et al., 2010). Furthermore, permits are often needed for catching bats, their handling comes with potential zoonotic risks (Wong et al., 2007), the animals become stressed and more vulnerable to predation (Rocha-Mendes & Bianconi, 2009), and can even succumb to this invasive sampling method.
Passive acoustic monitoring is also commonly used for sampling bats, since most bats vocalise in the ultrasonic range for navigation with so-called echolocation calls. Passive ultrasound recording relies on automated devices to record echolocation calls made by bats. Single, cheap devices can sample large spaces and be programmed to record for long durations. However, the vast majority of Pteropodidae, occurring in the Paleotropics and Oceania, do not echolocate (except genusRousettus ), which explains why capture-based methods are essential there. Still, little is known about bat acoustics in the tropics, and acoustic methods need to be adopted more widely, especially in the Paleotropics (Kingston, 2010). Also, bats do not necessarily have species-specific echolocation calls, and calls are variable (Obrist, 1995). As a result, many species cannot be distinguished on the basis of ultrasound alone and are grouped within ”sonotypes” (Walters et al., 2013). Finally, very high frequency bat calls usually attenuate quickly in air and are seldom picked up by microphones that have declining sensitivity with frequency. Some bats also produce narrow ultrasound beams which are less likely to hit a microphone (Brinkløv et al., 2011). Finally, sound detection spaces are species-specific and seldom accounted for (K. Darras et al., 2016). Thus, acoustic detection and identification of bats is challenging, and density estimation is nearly impossible - especially across species.
Mist-netting and passive acoustic monitoring are now established, standardized sampling methods for bat biodiversity surveys (Flaquer et al., 2007). It is often advised to combine both methods to reduce the overall sampling bias (Kuenzi & Morrison, 1998), especially where Pteropodidae occur. However, recently, a proof-of-concept has been proposed for technologically enhanced point counts to sample flying bats at night (K. Darras et al., 2021). These enhanced bat point counts are an active (i.e., requiring a human operator) sampling method to detect and identify all flying bats within a sampling area at night, combining thermal sensing to detect flying bats, ultrasound sampling to record their echolocation calls, and near-infrared imagery to capture their morphology. Thermal and near-infrared imagery have been used before to count bat colonies directly in caves (Betke et al., 2008; Sabol & Hudson, 1995), and thermal imaging has also been combined with ultrasound recording to detect bats with drones (Fu et al., 2018) and at wind farms (Correia et al., 2013). Near-infrared imaging can also detect pollinating bats (Frick et al., 2009). However, these studies surveyed sites with a great density of inactive bats, or focused on specific sites where a particular interaction occurs. Near-infrared imaging has not been used yet for identifying flying bats passively; it remains to be seen whether entire bat communities can be sampled with this method and how it compares to established methods.
Here, we showcase bat point counts and demonstrate how they can be used for ecological studies. We compare them against mist-netting and ultrasound recording in an agricultural system in the Paleotropics, where both insectivorous, echolocating bats and frugivorous, non-echolocating bats are common. We measure the detection spaces of all three sampling methods, present a novel, morphological-acoustic bat identification key tailored to our study system to make use of the acoustic and photographic data, investigate how accurately and efficiently the species pools are sampled by each method, and compare diversity patterns using rarefaction and extrapolation sampling curves. We discuss practical considerations, and we give an outlook as to the new possibilities offered by bat point counts for the study of bats.