Recently, machine learning has taken over glaciology as machine learning has been demonstrated to provide excellent classifications in remote sensing of glacier outlines (https://www.sciencedirect.com/science/article/pii/S0034425715301528), proved to be excellent physical simulators of glacier flows (https://www.cambridge.org/core/journals/journal-of-glaciology/article/deep-learning-speeds-up-ice-flow-modelling-by-several-orders-of-magnitude/748E962A103D2AF45F4CA8823C88C0C0), and has demonstrated to be adequate at selecting glaciers likely to have instabilities based on physical and climatic factors (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022JF006597).
This master's thesis will examine this last problem in detail. Recent work using data from Svalbard has been able to identify glaciers with instabilities that lead to surging behaviour (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022JF006597).
First to be examined are the glaciers that this machine learning model failed to classify correctly. What were the physical characteristics and climactic factors that the model missed? Based on the outcome of this study, the thesis will then build a new model that will be applied globally.