Three year to develop a snow avalanche detection service for the Austrian Alpine region using remote sensing data from Sentinel and automotive lidar data.

Each year about 100 people lose their lives in snow avalanches in the European Alps and in addition the annual financial loss due to road closures and damages is estimated to be more than one billion euros in Europe only. Avalanche forecasting is currently done by human experts. The overall avalanche activity of a specific region is often unknown even though this information is very important to deliver accurate predictions. Data from satellites can be used together with machine learning algorithms to provide insights on avalanche activities all over the Austrian Alpine region. Automatic avalanche detection using earth observation satellites enables a better and more complete mapping of avalanches than with manual reporting. This can be used to enhance avalanche prediction done by experts and to form the basis of research on automatic avalanche prediction methods. There is currently no operational snow avalanche detection service for the Austrian Alpine region available. Based on recent progress in satellite data resolution and frequency, some scientific studies are available focusing on some test regions in e.g., Norway and Switzerland. In the framework of this exploratory project, we plan to build the scientific basis for establishing a snow avalanche detection service for the Austrian Alpine region using remote sensing data. We plan to establish a strong consortium with background in remote sensing, reanalysis weather data, machine learning, and snow avalanche detection for a larger follow-up project. The exploratory project includes the collection of requirements and expectations related to the final avalanche detection/forecasting product from potential user groups and customers, as well as the development of a deep learning framework, a data pipeline, and an IT infrastructure that can process remote sensing and reanalysis weather data for detection of snow avalanches. We also plan a publication assessing the potential of a snow avalanche detection and forecasting system based on remote sensing and reanalysis weather data in the Austrian Alpine region.

Duration: 01.10.2021 - 28.02.2023

Role: Lead Software Architect and overall technical Lead


Thomas Goelles
Thomas Goelles
Scientist at Karl-Franzens-Universität Graz and Virtual Vehicle Research

My research interests include glaciology, lidar, albedo and numerical modelling of ice sheet processes