Wildlife monitoring is integral to species conservation, facilitating data-informed decisions (Conroy et al., Reference Conroy, Stodola and Cooper2012). Among other considerations, successful monitoring programmes must be cost-effective (Hinds, Reference Hinds1984). Given that conservation funding is limited and highly competitive, researchers must be able to maximize data collection efforts with minimal resources (Speaker et al., Reference Speaker, O’Donnell, Wittemyer, Bruyere, Loucks and Dancer2022).

Vocalizations are fundamental to the social lives of many animals and provide a wealth of data to inform research and conservation (Teixeira et al., Reference Teixeira, Maron and van Rensburg2019). Passive acoustic monitoring has become a standard monitoring method for nocturnal taxa, most notably bats, with ultrasonic detectors and autonomous recorders tracking activity, occupancy and diversity at landscape scales (Stilz & Schnitzler, Reference Stilz and Schnitzler2012; Sugai et al., Reference Sugai, Silva, Ribeiro and Llusia2019). Traditional methods of monitoring nocturnal and arboreal mammals often require direct observations or audible calls (Duckworth, Reference Duckworth1998). Passive acoustic monitoring is a viable tool for estimating occupancy and monitoring cryptic behavior in nocturnal species (Wrege et al., Reference Wrege, Rowland, Keen and Shiu2017; Martins et al., Reference Martins, Segurado and Marques2025). However, some mammalian species have evolved the ability to vocalize and hear at frequencies beyond the relatively limited audible range of humans (20 Hz–20 kHz), presenting monitoring challenges because of our inability to perceive ultrasound and the fact that high-frequency sound does not travel long distances before dissipating (Arch & Narins, Reference Arch and Narins2008). Attended recorders are effective for detecting ultrasound (Gursky-Doyen, Reference Gursky-Doyen2013), but there is a dearth of information on using autonomous recorders for terrestrial non-chiropteran mammals, indicating a need to investigate the potential of autonomous recorders to study ultrasound in these species (Sugai et al., Reference Sugai, Silva, Ribeiro and Llusia2019).

To expand passive acoustic monitoring, we need to determine which technology to use for an effective and cost-conscious monitoring programme. Two prominent brands in ultrasonic recording are Open Acoustics and Wildlife Acoustics, both based in the USA. For this study we compared the performance of one monitor from each company at substantially different price points. As of February 2026, Open Acoustics’ AudioMoth cost c. USD 89, whereas Wildlife Acoustics’ Song Meter Mini Bat 2 cost c. USD 749.

We conducted this study in June 2024 at the Sakaerat Slow Loris Project in north-east Thailand. We performed the experiments in three forested environments: dry dipterocarp, dry evergreen, and the ecotone between them. The habitats ranged in density from relatively open (dry dipterocarp) to moderately dense (ecotone) and dense (dry evergreen). We utilized comparable recording settings for the AudioMoth, with IPX7 waterproof case, and Song Meter Mini Bat 2, with IP67 waterproof case (Table 1; Kunberger & Long, Reference Kunberger and Long2023). We used the same devices across all trials to avoid any confounding factors, and implemented a frequency trigger to synchronize recordings with trial start times.

Table 1

AudioMoth and Song Meter Mini Bat 2 (SMMB2) recording settings for comparative experimental test of detection and clarity. Threshold level and window length are AudioMoth-specific parameters used by its frequency-trigger. The SMMB2 uses a triggered ultrasonic recording system configured via ‘minimum trigger frequency’ and ‘trigger window’, and does not provide user-adjustable threshold or window length settings, which are handled internally by the recorder.

We mounted the recorders 50 mm apart on a 2 m bamboo pole, and a portable loudspeaker (JBL GO3, JBL, China) on a separate, identical pole. We chose locations on established trails where characteristic vegetation was present, to ensure that foliage coverage was comparable across the three habitat types (Plate 1). We played modulated tones three times at six distances from the recorders (1, 2, 4, 8, 16 and 32 m) for 8 s during each trial during the night (0.00–3.00) and day (13.00–15.00) (Darras et al., Reference Darras, Pütz, Rembold and Tscharntke2016). We set the loudspeaker to 50% volume, then emitted modulated ultrasonic tones of 20–25 kHz (Sonic Pitch Sound Generator 3.3.0, Pixel Fox, China), always orienting the speaker towards the two monitors.

Plate 1

A daytime trial in a section of dry evergreen forest on an established trail. One person holds a 2 m pole (A) to which the loudspeaker (B) is secured, at a distance of 16 m (in this case) from a second person (C), who holds a 2 m pole to which the two ultrasonic recorders (AudioMoth and Song Meter Mini Bat 2) are secured.

We analysed recordings in Kaleidoscope 5.6.8 (Wildlife Acoustics, USA), visualizing them as spectrograms and assessing detection through manual inspection. For each detected tone, we drew a bounding box around the waveform on the spectrogram at 20–25 kHz to extract the mean decibel (dB) reading, which correlates to visual clarity on spectrograms (Plate 2), although it is not a true measure of clarity, such as signal-to-noise ratio. We used a Wilcoxon signed-rank test with a Bonferroni correction in R 4.4.2 (R Core Team, 2024) to determine if there were differences in dB readings between the recorders across the six conditions (each of the three forest types in the day and in the night). We used a Kruskal–Wallis rank sum test to compare the mean dB levels recorded across all combinations of location and distance. We assigned a value of −96 dB to the lowest decibel level on Kaleidoscope’s color contrast gradient, effectively indicating no detectable sound when no data were present on the spectrogram (Wildlife Acoustics, 2024).

Plate 2

The difference in visual clarity of sound (measured by decibels) recorded by (a) the AudioMoth and (b) the Song Meter Mini Bat 2 (SMMB2). The AudioMoth has greater visual clarity on the spectrogram than the SMMB2, as indicated by the intensity of the orange color.

The Song Meter Mini Bat 2 detected ultrasonic tones in 75% of trials but failed to detect ultrasound during the day in the dry dipterocarp forest at ≥ 16 m, during the day in the dry evergreen at ≥ 4 m, during the day in the ecotone at 32 m, and during the night in the dry dipterocarp and dry evergreen at 32 m. In contrast, the AudioMoth had a detection rate of 97%, only failing to detect ultrasonic tones in the dry dipterocarp and ecotone forests during the day at 32 m (Fig 1).

Fig. 1

Dumbbell plots of detection and decibel readings of modulated ultrasonic tones (20–25 kHz) for two ultrasonic recorders: the AudioMoth and Song Meter Mini Bat 2. Recordings were taken at six logarithmically increasing distances. The points are the mean decibel measurement of three trials at each distance for each recorder across three forest types (dry evergreen, dry dipterocarp and the ecotone between them) in daytime and night-time (a total of 108 measurements). Values closer to zero represent higher decibels. Absence of data at certain distances represents a failure to detect ultrasound.

Fig. 1 Long description

The image contains six panels of dumbbell plots showing detection and decibel readings of modulated ultrasonic tones for two ultrasonic recorders, AudioMoth and Song Meter Mini Bat 2, across three forest types: dry dipterocarp, dry evergreen, and ecotone. Each panel represents data collected during the day or night. The horizontal axis represents distance in meters, ranging from 1 to 32 meters, and the vertical axis represents decibel levels, with values closer to zero indicating higher decibels. Each point on the plots is the mean decibel measurement of three trials at each distance for each recorder. The green dots represent AudioMoth readings, and the orange dots represent Song Meter Mini Bat 2 readings. Absence of data at certain distances indicates a failure to detect ultrasound. Panel A: Daytime readings in dry dipterocarp forest. Panel B: Nighttime readings in dry dipterocarp forest. Panel C: Daytime readings in dry evergreen forest. Panel D: Nighttime readings in dry evergreen forest. Panel E: Daytime readings in ecotone. Panel F: Nighttime readings in ecotone.

It is well documented that ultrasound diminishes rapidly and reflects off substrates such as foliage (Stilz & Schnitzler, Reference Stilz and Schnitzler2012). However, even though the vegetation types in which we conducted these trials varied in density, there were no statistically significant differences in detected average dB levels at the same distances across habitats (Supplementary Table 1).

When comparing all dB means without considering conditions (i.e. distance, time of day and forest type), using the Wilcoxon test, we found no statistically significant differences between the recorders (Supplementary Table 2). The Kruskal–Wallis rank sum test revealed no statistically significant differences in dB levels between recorders across all combinations of location and distance (Supplementary Table 3). The mean dB differences were 1.08–14.2 dB, with the AudioMoth having higher dB readings in 25 of the 27 conditions in which both recorders detected ultrasound. Fig. 1 presents all comparisons across conditions.

The AudioMoth provided equivalent dB readings on spectrograms compared to the Song Meter Mini Bat 2 (Plate 2), and reliably detected ultrasonic tones from further away than the Song Meter Mini Bat 2, capturing ultrasound up to 32 m in most conditions. These findings indicate that the AudioMoth could be an effective and economical choice.

Our study is a small but illuminating trial that could help guide decision-making for research and conservation monitoring projects with ultrasound (Hinds, Reference Hinds1984; Teixeira et al., Reference Teixeira, Maron and van Rensburg2019). With improving knowledge of terrestrial non-chiropteran ultrasound, passive ultrasonic recording technology may become essential for monitoring nocturnal and cryptic species (Sugai et al., Reference Sugai, Silva, Ribeiro and Llusia2019). The narrow frequency band (20–25 kHz) and single deployment angle in this study limit the interpretation of these preliminary data, but the results suggest that the AudioMoth may offer a more cost-effective option for detecting modulated ultrasound at distances of up to 32 m. Additional experiments at higher frequencies and off-axis angles will clarify its performance under a greater range of field conditions. Because the Song Meter Mini Bat 2’s omnidirectional microphone samples a larger area than the directional AudioMoth, fewer Song Meter Mini Bat 2 units may provide comparable spatial coverage and require less effort, storage and analysis time. However, for the cost of a single Song Meter Mini Bat 2, a monitoring programme could purchase eight AudioMoths. Although this could expand geographical coverage, there are additional costs such as SD cards and costs associated with the effort to deploy each recorder. Users will need to evaluate this trade-off case-by-case to optimize data collection, storage and analysis (Hinds, Reference Hinds1984). Before selecting a recorder, users should also consult published microphone response curves to match device sensitivity to the focal species’ call frequencies.