From last in report in April we are now finalizing the paper for Modeling the Influence of Climate on American Avian Migration, studying how annual differences in temperature and precipitation across North and South America have historically affected avian migration timings in the Spring and Fall, and how this might change in the near term with climate change. We used machine learning and the Norwegian supercomputer Sigma2 on this study to preprocess hundreds of GB of eBird data.
We are also mid-way through paper studying how machine learning-derived soundscape features compare to climate features as covariates of avian species and of species richness using 6 years of continuous recording from a site in Ithaca, New York. We are also using Sigma2 on this project to derive soundscape features from many TB of audio data.
The candidate is also mid-way through aesthetics and philosophy paper where he is studying how digital art can extend the influence of scientific work and public conciousness related to the environment, especially in the context of using sound art. This includes a case study where he is using hundreds of bird calls derived from the Sound of Norway data to illustrate spatial and temporal changes in the Norwegian Forest Soundscape.
In the Sound of Norway project, the candidate has helped with experimental design, including identifying appropriate sites for setting up microphones, going out in the field and installing microphone/solar panel/battery systems, analyzing migration timings and species richness as a function of latitude, and studying how deployed recorders can supplement in-person bird surveys. This work will soon be published as a NINA report and a scientific paper next year.
In addition, the candidate has been involved in the Yellowstone snow scooter project, where he is analyzing density of bird calls before and after snowscooter detections to study how birds respond to and recover from noise pollution. This work will also be published as a NINA report, and a scientific paper next year.
The candidate has also assisted co-supervisor Sarab Sethi with a soon-to-be-submitted paper studying how machine learning-derived soundscapes features can be generalized across ecosystems and how much a change in soundscape represents a change in species richness. His role was deriving machine learning features and other acoustic indices from the 6 years of Cornell audio data on Sigma2.
At last, the candite is taking his last course at University of Oslo this Spring on Music and Machine Learning, and will have a course project, potentially focusing on audio style transfer, that can also become a publication.