website: https://github.com/ODINN-SciML/MassBalanceMachine
The researchers’ activities of “The Mass Balance Machine” are linked to those of IACS (International Association of Cryospheric Sciences).
__________________________________________________________________________________________________________________________
The Mass Balance Machine is a collaborative international effort. It’s part of the PhD projects of two students: Kamilla Hauknes Sjursen from the Western University of Norway, who focuses on its application in Norway, and Marijn van der Meer from ETH Zürich, who applies it to the glaciers of Switzerland. It is supported by Jordi Bolibar and Alban Gossard from the Univ. Grenoble Alpes and Prof. Harry Zekollari from the Vrije Universiteit Brussel. Two Master students, Michel Burlet and Julian Biesheuvel also worked on it.
The Mass Balance Machine aims to provide a data-driven alternative for modelling glacier surface mass balance, which refers to the change in mass at the surface of a glacier, or a part of a glacier, over a given period. It is a global machine learning framework that assimilates glaciological observations and remote sensing data to estimate glacier mass balance with high spatial and temporal resolution. Designed for regional-scale applications, it enables accurate reconstructions of glacier surface mass balance across diverse settings.
The main scientific questions of this project are:
- How does the Mass Balance Machine compare to traditional physical models of glacier mass balance?Traditional physical models, such as temperature-index or energy-balance models, rely on simplified physical relationships and often require extensive calibration using in situ data, which can be a limitation at larger scales. In contrast, the Mass Balance Machine is a data-driven machine learning model that learns statistical relationships directly from available glaciological and remote sensing data. This allows it to more flexibly integrate diverse data sources and better capture regional variability, especially in cases where direct observations are sparse or unavailable. MBM has shown improved performance in estimating seasonal mass balance compared to several established physical models, particularly in heterogeneous climate settings.
- What are the Mass Balance Machine’s capabilities to extrapolate to glaciers without in situ measurements and other periods in time?
One of the strengths of the Mass Balance Machine is its ability to generalise beyond the glaciers and periods used in training. By learning from a wide range of climate and glaciological conditions, MBM can make reliable estimates of mass balance for unmonitored glaciers and in time periods where no direct measurements exist. This extrapolation capability has been validated in Norway and is currently being tested in other regions, such as Switzerland, to assess how well it transfers across geographical settings and climate regimes.
Why is this research important?
Glacier evolution models based on temperature-index approaches are commonly used to assess hydrological impacts of glacier changes. But in in large-scale applications, these models lack calibration frameworks that efficiently leverage sparse high-resolution observations. Machine learning approaches can potentially address this limitation by learning relationships from sparse data that are transferable in space and time, including to unmonitored glaciers
What is one key discovery and its challenges?
In a recent study on Norwegian glaciers, we discovered that machine learning models like the Mass Balance Machine (MBM) can generalise across glaciers and climatic settings using relatively sparse data, outperforming traditional glacier evolution models—particularly in reconstructing seasonal mass balance with significantly lower error. This highlights the potential of data-driven approaches to improve regional-scale glacier mass balance modelling where observational data are limited. A key challenge remains in assessing how well such models can generalise to entirely different geographical regions, such as other parts of Europe or the world, which is currently being investigated.
Read more: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1206/
Don’t miss next week’s blog posts featuring interviews with ECS participating in “The Mass Balance Machine” project!