The primary goal of this workshop is to work on a project with one of the instructors for a week. There are three options for participants: you can either join one of the two pre-established projects, meant to learn the basics of machine learning applied to glacier modelling or remote sensing, or you can bring your own project.

[ Project 1 ] - Modelling glacier mass balance with machine learning - Jordi Bolibar, Kamilla Hauknes Sjursen and Facu Sapienza

The goal of this project will be to learn and apply different types of machine learning and statistical regression methods (e.g. Lasso/Ridge, Random Forests, Gradient Boosting, neural networks...) to simulate glacier mass balance (MB) for all glaciers in Scandinavia. We will use the Open Global Glacier Model (OGGM) to retrieve glaciological and geodetic MB observations, and we will train multiple ML models using topographical and climate data. By using a common set of cross-validation folds, we will compare the different ML methods in order to understand their strenghts and weaknesses. Moreover, we will learn how to add soft physical constraints to the models. If needed, the project could be extended to also include ice flow dynamics, and eventually physics informed learning.




[ Project 2 ] - Glacier mapping from optical and SAR satellite imagery using Deep Learning and Random Forests - Konstantin Maslov and Thomas Schellenberger

The students will have the possibility to learn how to map glaciers in Svalbard, the Alps and Pakistan from Sentinel-1, Sentinel-2 data and a digital elevation model using fully convolutional neural networks and random forests. They can also test the scalability and transferability of several networks to different regions and years or extend the exercise to map glacier surface types.




[ Project X ] - Bring your own project

Bringing your own project requires planning and an achievable goal for the workshop. In order to do so, we would ask the participant to share a short explanation of the project, its current status and the objectives for the workshop in the application form. Teaming up with other participants for a project is highly recommended, in order to foster interactions and create collaborations.