Coupled Snow-Runoff Modelling in Alpine Karst: Comparing physics-based vs. conceptual snow model representation in neighbouring catchments with different characteristics

Hydrological modelling in snow-dominated high-alpine karstified catchments remains challenging due to complex snow processes and their influence on runoff generation. This study investigates the impact of snow model complexity on discharge simulations in the Zugspitze region, at the border of Germany and Austria, in the European Alps across two neighbouring catchments. Both catchments are located in the same mountain range, are heavily karstified, share similar geological structures, and have practically no surface runoff. However, some catchment characteristics differ, which we will investigate in this modelling study. The Partnach spring catchment (15.4 km², 1430-2962 m a.s.l., W-E orientation) comprises steep rocky terrain in the upper and lower part in the south-, west- and north-facing terrain with dominant a high-altitude plateau in the middle part, and overall limited vegetation. In contrast, Hammersbach (17.8 km², 768-2951 m a.s.l., N-S orientation) shows stronger elevation gradients and is predominantly covered by forests in the lower third of the catchment. Steep north-facing rock walls reach down to ~1000 m and lead to a longer lasting snow cover in the upper and middle part of the catchment due to terrain shading. We examine for these two catchments how a physically-based snow representation (Alpine3D) compares to a degree-day approach (CemaNeige) regarding its impact on snowpack evolution and runoff generation when coupled with an identical lumped conceptual GR4H hourly routing scheme and meteorological forcing. Alpine3D explicitly simulates boundary layer fluxes and energy balance processes, while CemaNeige relies on the upper GR4H storage to represent evapotranspiration and interception. The GR4H routing parameters are calibrated using a 5-year moving window approach across hydrological years 2014-2025 for both catchments, which face characteristic high-alpine measurement challenges such as winter data gaps, avalanche events, and sediment transport. Multicriteria validation incorporates SWE measurements, Sentinel-2 snow-covered area information, and discharge observations. Results show on the one hand, that both modelling approaches achieve comparable annual discharge performance, while Alpine3D consistently provides a more realistic representation of spatiotemporal snow distribution. On the other hand, the comparative analyses of the adjacent catchments present model behaviour under different snow, terrain and land-cover conditions. This study provides insights into the conditions under which increased physical realism improves runoff simulations and for which situations conceptual approaches are a good choice, supporting informed model selection for snow-dominated and ungauged alpine regions.