Glacier mass balance models, accurate and robust on a global scale, are of particular importance to climate scientist to analyze interactions between the earth’s cryosphere and climate and ultimately predict sea level rise. Rare in-situ measurements on glaciers complicate the development of respective models that need to work globally in a variety of climate settings. Alongside most common temperature and precipitation index models this paper investigates the utilization of artificial neural networks for the prediction of glacier mass balances on different scales. The estimation of annual WGMS data by a neural network trained on monthly averaged temperature and precipitation data provides promising results and significantly outperforms corresponding index models. In contrast, more investigations are needed to verify the robustness of neural networks, when it comes to the approximation of surface height changes of glaciers based on AWS data on an hourly scale.
