In this project, we are developing novel machine learning approaches to analyze recordings of bat calls and leverage the calls as a proxy to identify environmental problems. Animal observations are already being used to identify environmental issues. However, bat calls have not been used systematically for this task, yet can strongly contribute in this area.
In an initial analysis of bat recordings, we were able to show that it is possible to identify groups of similar bat calls using machine learning approaches. We hypothesize that we can identify specific types of bats or individual bats, as well as behaviour based on the calls using novel machine learning approaches.
Our main objective is to develop new machine learning methods to find patterns in collections of bat calls and relate these call patterns to patterns in environmental data. We will use water pollution data as a starting point, taking advantage of the direct connection to bats – certain pollution will impact the population of food sources for bats which in turn will cause bats to migrate and change their behaviour.