Savannas are a major terrestrial biome that cover about one fifth of the global land surface. The world's savannas have been extensively transformed by human activities including conversion to rangelands for livestock grazing and clearing for cropping. Despite the savannas of northern Australia being the largest and least cleared on earth, significant population declines have been observed among some vertebrate fauna, with declines particularly evident among ground-foraging granivorous birds. The southern black-throated finch (SBTF; Poephila cincta cincta) is one such species whose range has contracted by over 80% from its historical distribution. Now listed as Endangered, the species has a present-day distribution restricted to two population strongholds. Most previous research on the species has focused on the species' population in the Townsville Plains Bioregion, despite the less-researched population in the Desert Uplands Bioregion occurring in a substantially more arid area, with less surface water availability, limited overlap in vegetation communities and less habitat fragmentation. These climatic and habitat differences are likely to drive different patterns of movement and habitat use between the populations. Understanding these patterns is important for the targeted conservation of SBTF in the Desert Uplands Bioregion. However, monitoring rare and threatened species over large areas is challenging and this thesis explores avenues of methodological improvement that are broadly applicable. Methods to investigate avian movement ecology have rapidly progressed in recent decades, driven by advances in sensor technologies. Prominent among these are automated radio telemetry (ART) systems and acoustic recorders. Both methods produce large volumes of data, necessitating continued development of analytical tools to extract meaningful ecological insights. For ART systems, the primary challenge is generating accurate location estimates from complex signal data. While for acoustic recorders, a key challenge is developing deep learning call detection models for species that are rare, difficult to record and lack pre-existing training datasets. This thesis aims to develop methods to address these methodological challenges that have broad relevance in movement ecology research. I then apply these methods to investigate the movement ecology of the SBTF population within the Desert Uplands Bioregion. In response to the challenges of localising wildlife positions from complex ART signal data, I introduce a novel machine learning-based location fingerprinting method for ART systems and compare its performance with two alternative methods, biangulation and linear regression. The location fingerprinting method achieved slightly better accuracy than the linear regression method and substantially outperformed the biangulation method. Broadly, the location fingerprinting method provides a greater versatility of use than alternative methods, allowing application across various ART system designs with minimal need for study-specific customisation. To improve methods to build deep learning models to detect rare bird calls, I implement an active learning framework, informed by the broader machine learning literature. This framework uses a human-machine interface to selectively label data that are most likely to improve model performance. I demonstrate that this active learning framework is an efficient strategy to create a state-of-the-art call detection model in a scenario with no pre-existing training data and where target calls are extremely rare within a large, unlabelled dataset of long-term audio recordings. I then apply the ART system and bioacoustic methodological developments from this thesis to investigate the seasonal habitat use, movement and activity patterns of SBTF within the Desert Uplands Bioregion of Queensland, Australia. My results show that over short periods (of less than one month), SBTF are largely sedentary but occupy extensive home ranges. Within these ranges, they make daily movements among foraging, drinking and nesting sites. Few landscape-scale movements exceeding 4.5 km were observed. I found that the distances SBTF travel from permanent water sources were substantially larger than previously reported for the Townsville Plains SBTF population. I recorded an average resighting distance of 1.2 km from water, with some observations as far as 4.2 km from water, which was close to the maximum distance from water that could be achieved within the study area. Additionally, both telemetry and bioacoustic studies found that SBTF diversified their habitat use in the dry season compared to the wet season, potentially due to SBTF using a broader area of habitat when food resources were relatively scarce. The methodological developments presented in this thesis introduce new open access tools to improve ART localisation methods and facilitate the creation of call detection models for rare and difficult to record species. These tools establish a foundation for refining location fingerprinting techniques in wildlife telemetry and developing reliable methods to monitor SBTF population sizes using acoustic recorders. The insights into movement ecology derived from this research can be used to inform conservation strategies for SBTF and have broader implications for the monitoring and management of species across large spatial scales in environments with substantial spatiotemporal variation in resource availability.