To achieve our objectives, the project was divided into two major components. For the first component of the project, we tested various algorithms and methods for classifying mating behaviours from accelerometry, using captive male ruffs (Calidris pugnax). For the second component of the project, we recorded and analysed the behaviour and reproductive success of pectoral sandpipers in their natural habitat.
Ruffs are a polymorphic, lekking shorebird with highly skewed mating success. We fitted 20 captive males with accelerometers and recorded their behaviour using video cameras. We then trained machine learning algorithms to identify behaviours from accelerometry. We compared the classification performance of three supervised machine learning methods: Random Forest, Hidden Markov Models and Neural Networks. During this process, we also identified issues that can cause overestimation of model performance. A full description of the methods and results, including the associated code, is being prepared for publication. Our findings will also be presented at the European Ornithologists’ Union conference (Lund, Sweden) in August 2023.
Like many other projects, our research was disrupted by the COVID-19 pandemic. We were unable to travel to our field site in 2021, which caused some delays to the field component of our study. In 2022, we fitted 100 male pectoral sandpipers with accelerometers at our study site near Utqiagvik, Alaska. As with the ruffs, we also recorded male behaviour using video cameras. To determine reproductive success, we caught adults and searched nests throughout the study site and collected a small blood sample from all males and females, as well as from all offspring. Our analysis is still ongoing, but so far our results replicate a previous study that found that more active (sleepless) males sired more offspring. We are now using the accelerometry data and machine learning methods tested with the ruffs to quantify how much time and energy the males spent on various courtship and territorial behaviours. We expect that these results will be published and made publicly available later this year.