Integrating snow-water equivalent simulated by a physically based model into a lumped model in an Alpine catchment in Italy

In the European Alps, seasonal snow plays a crucial role in hydrology, functioning as a reservoir by storing precipitation during winter and releasing it during the summer. Snow is highly sensitive to climate change, particularly in low- and mid-elevation mountain regions like the European Alps. In snow-fed basins, any changes in snowmelt contribution to river discharge can significantly impact agriculture, domestic water supply, and hydro power generation. 

Hydrological modeling employs a variety of models, ranging from simple lumped models to physically-based, spatially distributed models, to simulate river discharge. These models either have a simple temperature-based or a physically based snow module to simulate the snow dynamics. Distributed, physically based models can provide accurate insights into snow dynamics. However, their high input data requirement, over-parameterization, and high computational demands make them challenging to calibrate for discharge estimation for operational purposes. In contrast, simple lumped models require less input data, standard snow parameters, quick calibration, and are well-suited for operational applications, but, of course, lack spatial details.

In this study, we present an approach to improve both runoff forecasting and spatial snow pattern estimation by integrating the snow water equivalent (SWE) simulations from a physically based GEOtop model into the lumped GEOframe system. We utilize a mass-conserving Topographic Response Unit (TRU) aggregation logic to preserve the spatial variability of melt fluxes across elevation and aspect gradients. The methodology is applied in the Non Valley catchment, Italy, where water is important for agriculture, hydropower, and other uses.

Our results for the period 01-01-2017 to 15-09-2022 at hourly time step show that the GEOframe is able to simulate the discharge very well with a Kling-Gupta Efficiency (KGE) value of 0.87 and 0.72 during the calibration and validation, respectively. Substituting the internal snow module with GEOtop-derived fluxes yielded a KGE of 0.71 without further calibration. This demonstrates that the physically-based snow input successfully maintains the model’s predictive power while providing a more realistic and spatially distributed representation of snow dynamics. This coupling approach preserves the operational efficiency of lumped models while incorporating the improved physical representation and spatial variability essential for modeling mountain hydrology under a changing climate.

Acknowledgement

JMW and RR would like to thank and acknowledge the funding support from Project “SPACE IT UP! ASI Contract n.2024-5-E.0 CUP Master n. I53D24000060005” SAP fund n: 000040104905.