Nest site selection is the principal way secondary cavity-nesting species mitigate the negative effects of factors such as predation, parasitism and exposure on productivity. High-quality cavities could then be expected to be selected in response to the primary threat to nest success. Understanding how demographic rates are affected by anthropogenic changes to ecosystems is vital if effective conservation management strategies are to be developed and implemented. Large-bodied secondary cavity-nesting birds rely on large cavities in mature trees that are often absent or reduced in anthropogenically disturbed forests. Thus, the availability of high-quality nest sites may be limited for these species, potentially reducing productivity. The aim of this study was to investigate nest-site selection and the effect of nest-site features on productivity in the critically endangered great green macaw (Ara ambiguus) in northern Costa Rica. We show that great green macaws select nest sites according to the characteristics of the cavity and of the tree in which they are located. Moreover, productivity was a function of certain cavity features. We conclude that great green macaws are not reliant on primary forest for nest sites and typically choose cavities in remnant, isolated trees in pasture or young secondary forests.

Passive acoustic monitoring (PAM) - the use of autonomous recording units to record ambient sound - offers the potential to dramatically increase the scale and robustness of species monitoring in rainforest ecosystems. PAM generates large volumes of data that require automated methods of target species detection. Species-specific recognisers, which often use supervised machine learning, can achieve this goal. However, they require a large training dataset of both target and non-target signals, which is time-consuming and challenging to create. Unfortunately, very little information about creating training datasets for supervised machine learning recognisers is available, especially for tropical ecosystems. Here we show an iterative approach to creating a training dataset that improved recogniser precision from 0.12 to 0.55. By sampling background noise using an initial small recogniser, we can address one of the significant challenges of training dataset creation in acoustically diverse environments. Our work demonstrates that recognisers will likely fail in real-world settings unless the training dataset size is large enough and sufficiently representative of the ambient soundscape. We outline a simple workflow that can provide users with an accessible way to create a species-specific PAM recogniser that addresses these issues for tropical rainforest environments. Our work provides important lessons for PAM practitioners wanting to develop species-specific recognisers for acoustically diverse ecosystems.

Most conservation relies on being able to estimate population size accurately. The development, implementation and adaptation of effective conservation strategies rely on quantifying the impacts of different threats on population dynamics, identifying species that need conservation management, and providing feedback on the effectiveness of any management actions. However, current approaches are not suitable for wide-ranging species that reside in tropical ecosystems. Here we use the great green macaw Ara ambiguus as a case study to show that passive acoustic monitoring is an effective tool for collecting data that can then estimate abundance. We estimate a population of 485.65 +/- 61 SE great green macaws in Costa Rica during the breeding, suggesting the population here is larger than previously estimated. We have also highlighted potentially important areas for the species in regions that had not previously been studied. We have demonstrated at a population scale that passive acoustic monitoring (PAM) offers conservationists an efficient and effective way to understand population dynamics. With a high proportion of parrot species threatened globally, passive acoustic monitoring will enable effective monitoring and become an essential tool in conservation planning and evaluation. PAM technology has enormous potential to facilitate such assessments because it is easily scalable, recordings can be stored and re-analysed as machine learning, and abundance estimation techniques become more advanced.