Bioacoustic monitoring has great potential for accurate population monitoring over large spatial and time scale. However, it remains a challenge to efficiently analyse the large amount of data that can be collected in the field. This project will look at testing and developing new machine learning methods to (i) preprocess recorded sounds, (ii) extract the most relevant features and (iii) efficiently cluster the bioacoustic signals. These methods will be then applied to an available dataset of recordings from seabird colonies. Because these species spend most of their time at sea and are only found on land during some time, they are particularly difficult to monitor and little is known about their biology. Moreover, methods will be developped to infer the dynamic of remote population from bioacoustic data.