Comparing point counts, passive acoustic monitoring, citizen science and machine learning for bird species monitoring in the Mount Kenya ecosystem.

Biodiversity loss is a pressing challenge, with ecosystems across the world under threat from factors such as human encroachment, over exploitation and climate change. It is important to increase ecosystem monitoring efforts to provide actionable insights for ecosystem managers and to allow effective use of conservation resources. In this work, we compare traditional bird survey approaches using point counts with the use of autonomous recording units and citizen scientists' data at two sites within the Mount Kenya ecosystem. We also present a new dataset of more than 20 h of recordings obtained from the Mount Kenya ecosystem and annotated by expert ornithologists, and investigate the use of large deep learning models to process these recordings. Our results are mixed, and at one site, autonomous recording units and traditional point counts yield similar conclusions when comparing relative abundance of species, while at the second site, conclusions differ. Our results indicate that citizen science is preferable to point counts and autonomous recording units in determining species lists for particular habitats. However, even with the use of multiple methods, our survey still misses rare species known to occur in the Mount Kenya ecosystem, indicating that even the use of multiple methods is not exhaustive.This article is part of the theme issue 'Acoustic monitoring for tropical ecology and conservation'.