Results

Detected species

We found 100 thermal detections of seven taxa in point counts, 2009 detections of six taxa in automated ultrasonic recordings, and captured 83 bats from seven species in mist nets (Fig. 3). We excluded eight point count detections where neither ultrasound nor near-infrared pictures were recorded. We found one BFM, three CF, and three FM-QCF sonotypes (Table S1, Figure S1). We identified the FM sonotype to genus in automated ultrasound recordings, but found its putative identity from mist-netting data (Kerivoula pellucida ). We identified all CF sonotypes to species (Hipposideros kunzi, H. orbiculus, Rhinolophus sedulus ), and none was found using mist nets. All FM-QCF sonotypes were found using bat point counts and automated ultrasound recordings. One FM-QCF was identified to species using acoustic data (Pipistrellus stenopterus , (Kingston, 2013)) and was not found in mist nets. Using relative measurements from near-infrared imagery, one FM-QCF sonotype consisting of two candidate species was resolved to species-level in point counts (Scotophilus kuhlii ); it was the only species detected by all methods. The third FM-QCF sonotype was a complex of six candidate species and was reduced to three candidate species using near-infrared imagery. It was putatively identified with mist-netting (undescribed Myotis sp.1sensu (Huang et al., 2014)). One pteropodid genus (Cynopterus sp.) was detected in bat point counts and resolved to three distinct species in the mist-netting dataset (Cynopterus sphinx, C. brachyotis, C. minutus ). Mist nets detected one pteropodid species from another genus (Macroglossus minimus ) and conversely, aEonycteris/Rousettus genera complex was detected in bat point counts.

Rarefaction and extrapolation sampling curves

At 95 % sampling coverage, with incidence-based data, bat point counts detected a higher mean projected species richness, earlier than other methods; with abundance-based data, bat point counts reached a higher mean projected species richness, but at higher numbers of individuals (Fig. 2). At low numbers of sampling hours (≲ 2.5) and of sampled individuals (≲ 20), bat point counts and automated recordings allowed to detect significantly more species than mist-netting. The abundance-based extrapolation curves saturated more quickly for ultrasound recording and mist netting than for bat point counts, indicating that bat point counts had a higher probability to detect new species with increasing numbers of individuals. We provide an in-depth analysis of rarefaction-extrapolation sampling curves (Text S3, Fig S2).

Acoustic and thermal detection spaces

Bat point counts swept a large thermal detection area that encompassed a larger area than our mist nets, and their ultrasound detection spaces were larger and more narrow than those of the automated ultrasound recorders (Fig. 1). Ultrasound detection ranges of bat point counts (where the microphone was fitted with a horn) were almost three times larger than the unattended ultrasound recorders’ ranges (without horn) in the direction the microphone was pointing to (bat point counts: 450 m; automated ultrasound recorders: 164 m), and to some degree also to the side (Figure S3). The thermal scope had a range of 48 m on average, with a minimum of 19 m to a maximum of 84 m; its range was usually limited by obstacles such as oil palms or terrain irregularities. The mist nets approximately delimited an area of 150 m2when a quadrilateral was drawn across their outer corners.