Using bioacoustics and direct monitoring to improve the scale and quality of threatened species assessments in the tropics

Effective conservation requires effective monitoring. Rapid changes to ecosystems and increasingly extreme climate mean large-scale, robust data are needed to develop feedback into conservation strategies. Two factors critical to developing and maintaining conservation strategies are monitoring distribution, status and productivity. In tropical regions monitoring these factors are particularly difficult as the physical properties of the ecosystem can limit accessibility and the scale at which traditional monitoring can be carried out. Wide-ranging, grouping living species such as parrots are even more challenging to robustly monitor as they are canopy dwelling and violate key assumptions of standard survey techniques. If we are to conserve species like this and their ecosystems, we need to take advantage of significant increases in capacity and reductions in the cost of cutting-edge technology such as bioacoustics and drones. Here, I show, using the great green macaw (Ara ambiguus) as a case study, that passive acoustic monitoring is an effective tool for monitoring parrots and other challenging to study species like it. I developed a recogniser that consists of a two-stage machine learning pipeline to extract target species calls. This data was then used to estimate abundance and found that there are 485.61+/- 65 great green macaws in Costa Rica. This is a significant improvement on previous estimates based on extended point counts. Passive acoustic monitoring derived data was also used to model spatio-temporal distribution of the great green macaw across its range in Costa Rica. Both passive acoustic monitoring studies highlighted potentially important regions for the species in previously unstudied areas, demonstrating the value of passive acoustic monitoring as a conservation science tool. I then demonstrated how traditional approaches to monitoring productivity can be combined with drones to scale data collection. This study found that the productivity of great green macaws is within the range of other large macaw species (1.33 chicks per breeding attempt). They also select deep cavities in isolated trees that passively reduce the risk from arboreal and avian predators. Results suggest that avian predators are the primary cause of productivity loss in the study population. Importantly for the conservation of the species, results indicate that restoration of pasture or scrubland around nest sites could increase predation of nests by increasing accessibility to non-volant predators. My work demonstrates how technology can be a valuable tool for conservation science by increasing the potential scale and reducing fieldwork effort and resource expenditure. Furthermore, this work provides critical insights into the status and distribution of the great green macaw in Costa Rica. The insights gained and the approach demonstrated here are highly applicable to other parrots and species like them.