Disturbance of primary forest through selective logging and forest fire affects large areas of Amazonia and can substantially impact avian communities through alteration of forest structure and vegetative composition. However species specific responses to disturbance can often be highly idiosyncratic, making it difficult to infer the responses of poorly-known species. Nocturnal bird species are one such understudied group, which is particularly problematic as it has been argued that they may have disproportionately important ecological roles and be particularly vulnerable to local extinction.
Passive acoustic monitoring techniques facilitate large-scale study of nocturnal species, by removing logistical constraints for long duration surveys without affecting behaviour. Processing of the resultant audio data to species level can be done with automated classification models to efficiently analyse large data collections, but avoiding high levels of false positives and biases in error remains challenging. We present a method using open-source acoustic classification toolbox Tadarida to accurately detect and classify sound events with a low false positive rate, and a second classification stage to remove heterogeneity of error. Using this method, we find find that undisturbed primary forest has relatively low nocturnal species richness, none of the target species were most commonly detected in undisturbed forest, and that all of the nocturnal avian species appear to be robust to at least some degree of disturbance, with many persisting even in heavily disturbed forest.
Register for the event here. A Zoom link will be sent to you shortly before the event.
Registration will close on the 6th of April 2021 at 23:30.