Extracting Information from Environmental Data

Environmental science benefits from in large-scale and long-term environmental monitoring. Our wireless sensor networks are deployed in extreme high alpine regions. Seismic emission sensing using geophones opens up a new perspective on geophysical phenomena in the alpine environment. However such sensors pose enormous challenges in terms of energy consumption, computing power and communication. We work on state of the art algorithms in the context of environmental monitoring while reducing their computational footprint. Our results focus on the local pre-processing of data at the edge using machine learning and the analysis of data from different sensor modalities (images, weather data, seismic signals) using representation learning methods. The focus of these methods is on the classification of seismic data into classes that are relevant for geophysical research and early warning scenarios and those that are not.