The use of acoustics for monitoring tropical bats in Southeast Asia

Halting the loss of biodiversity is one of our century's greatest challenges. In Southeast Asia, the biodiversity crisis is driven by unprecedented rates of forest clearance and degradation. Here, conservation efforts face the inherent complexity of trying to protect species in landscapes that also support human livelihoods. Most of our understanding of how bats are affected by land-use change in the tropics is limited to those species which can be monitored using live-trapping techniques. However, acoustic monitoring is an important survey method for monitoring the whole bat community in other regions of the world. In this thesis, advanced technological and statistical approaches are used to investigate how tropical bats respond to land-use change in Borneo. Specifically, the thesis explores the application of acoustic monitoring for bats in this region.

In the first chapter, I outline how conservation zoning can be effective at protecting tropical bat diversity. To do so, I use a combination of live-trapping and acoustic monitoring to assess bat diversity and activity within The Crocker Range Biosphere Reserve compared to the surrounding agriculture. This includes demonstrating how acoustic data can be used to analyse differences in bat activity between habitats, even if it is not possible to classify calls to sonotype/species. Namely, this research highlights the importance of strict protective zones for conserving forest specialist species.

Next, I refine the bat call classification process by using acoustic data manually classified sonotype/species to assess the value of conservation set-asides. These data were combined with forest structural metrics measured using airborne LiDAR to assess the effectiveness of riparian reserves (areas of native forest along waterways) for conserving bats in oil-palm landscapes. Using Bayesian occupancy modelling, I demonstrate how - unlike other components of tropical biodiversity - forest quality is more important than riparian reserve width for maintaining suitable habitat for bats. This provides important evidence for designing effective conservation policies for tropical mammals. The lack of reference call libraries and automated classification tools in Southeast Asia impedes ongoing acoustic monitoring for bats in this region. Acoustic monitoring generates large datasets - often approaching the scale of big data - which requires time and expertise to manually process. These costs limit the application of bioacoustics as a feasible method for monitoring bats in tropical regions, not least in Southeast Asia - thus, limiting our understanding of how bat communities persist in human-modified landscapes. Therefore, for the third data chapter, I focus on developing a semi-automated framework for classifying bat calls in Southeast Asia when reference libraries may be limited. As proof of concept, this viii framework is then used to develop a classifier for the bats of Borneo using reference calls for 52% of all 81 known echolocating species on the island. This classifier was developed using free software to ensure the same framework could be applied to other regions of Southeast Asia.

Last, I apply the newly-developed classifier to examine patterns of bat activity in response to habitat disturbance at the Stability of Altered Forest Ecosystems Project in Sabah. The data used in this assessment represented a seven-fold increase in the number of recording hours compared to the combined effort of the previous studies. This acoustic dataset is combined with forest structural metrics measured using airborne LiDAR to assess responses across a gradient of habitat disturbance - comprising old-growth forest, logged forest, and monoculture tree plantations. In this chapter, I demonstrate how logged forests can provide important habitat for bats in human-modified landscapes, maintaining acoustic diversity and activity of many common species, particularly when compared to tree plantations. However, I also demonstrate how old-growth forest remains important for conserving forest specialist species.

Collectively, in this thesis, I document how bats can benefit from conservation initiatives that protect landscape features within human-modified landscapes. Crucially however, I also demonstrate that as disturbance intensity increases, less resilient species are lost from humanmodified landscapes. Whilst I provide important empirical contributions for understanding bat responses to disturbance, there remain substantial questions about how best to conserve these species and promote sustainable land use. The pipeline and classifier provided in this thesis will help improve accessibility for future acoustic studies to address these questions in Borneo and other areas of Southeast Asia. Therefore, the findings and resources presented here are an important step towards evidence-based land management for conserving Southeast Asian bats in human-modified landscapes.