Exploring the potential of aerial drone imagery to distinguish breeding Adélie (Pygoscelis adeliae), chinstrap (Pygoscelis antarcticus) and gentoo (Pygoscelis papua) penguins in Antarctica

• Differences in phenotype, posture, guano, and spatial distribution are analysed. • Most distinguishing features can only be detected at certain times. • Distinction based on phenotype is most reliable but requires high-res images. • In low-res images, distinction based on guano and spatial distribution is possible. • Green guano occurred regularly in breeding Adélie penguins early in season. Penguins are critical indicator species for detecting changes in the Antarctic ecosystem. As climate change impacts the Antarctic environment, the distribution of penguin breeding populations is also shifting. In the Antarctic Peninsula region, several penguin species’ breeding areas overlap. Given the isolated location of penguin colonies, remote sensing data are essential to detect large-scale changes in their distribution and species composition. While it is established that remote sensing data can effectively monitor penguin breeding populations during the breeding season, species differentiation remains challenging due to their similar appearance. The aim of this study is to explore the possibilities and limitations of species discrimination in aerial optical imagery, particularly using very high-resolution aerial imagery. We investigate the spatial resolution required to distinguish species based on their phenotype. Additionally, we explore the possibility and reliability of discriminating species by analysing differences in guano cover appearance, spatial distribution of individuals, posture, and the presence of chicks. Our findings indicate that differentiation is feasible for nearly all the investigated distinguishing features, although the reliability and spatial resolution requirements of the aerial images vary significantly. Consequently, these results enable specific flight planning for optimal species discrimination under the given conditions and serve as the basis for future automated mapping of penguin species using machine learning, facilitating early detection of changes.