Predicting Surface Mass Balance of Valley Glacier using Machine Learning

The retreat of glaciers has received considerable attention due to its implications for water availability and hydropower generation, thereby raising significant concerns for both the environment and society. Consequently, understanding the impact of climate on glacier evolution has become essential. In the present study, we investigate the application of various Machine Learning/Deep Learning models, specifically Linear Regression, Neural Networks, XG Boost, and Random Forest, to predict surface mass balance across two geographically distinct regions: the Swiss Alps and Svalbard. We also compared and analyzed different input datasets, such as ERA5, ERA5-Land (higher resolution), and a downscaled climate dataset to understand the impact of selecting different climate datasets and spatial resolutions. The performance of these models is evaluated based on different combinations of input variables to ascertain their impact on prediction accuracy.