Surface mass balance modelling of the Alps constrained by geodetic and snow line observations

Modelling the evolution of mountain glaciers in response to climate change is essential for accurate projections of global sea level rise and changes in the regional hydrological cycle. Glacier evolution and Earth system models applied at regional to global scales typically rely on simple temperature-index and snow accumulation models to describe spatio-temporal variations in glacier surface mass balance. The major advantage of temperature-index models over more complex energy balance models is their computational efficiency, the limited number of calibration parameters and the global availability of the required basic input data (i.e. air temperature and precipitation). However, temperature-index models based solely on an empirical relationship between melt and air temperature are not suitable for tropical and subtropical regions where incoming shortwave radiation and evaporation have a major control on the energy exchange at the glacier surface. In addition, several studies have shown that these models are oversensitive to air temperature variations and are not robust over time, so they need to be recalibrated for changing climatic conditions. This is problematic for forward modelling. Models of intermediate complexity, such as simplified energy balance models, are thought to be more robust over time and therefore more suitable for long-term modelling. The drawback of more complex models, however, is that they are more computationally expensive, require additional input data and have more degrees of freedom, making them prone to equifinality problems. Most glacier surface mass/energy balance models are now calibrated against geodetic observations available for basically any glacier worldwide. While these observations provide a consistent basis for model calibration, they do not allow the mass balance gradients to be constrained. This uncertainty can lead to large over- or underestimates of ablation and accumulation rates, with consequences for modelling glacier runoff and evolution. Here we present the results of a modelling experiment in which we compared the robustness and spatio-temporal transferability of two surface mass balance models of different complexity, constrained not only with geodetic but also with snowline observations automatically derived from Sentinel-2 data, for all monitored glaciers in the Alps.