Rain-on-snow (ROS) events cause repeated flooding in many mountainous regions with a seasonal snow cover. The complex interaction of processes across spatial scales makes it difficult to accurately predict the effect of snow cover on runoff formation for an upcoming ROS event, often resulting in underestimating the flooding potential of a respective event. To improve predictability of such events, the present study aims to identify the dominant snowpack runoff formation processes at different spatial scales. Snow cover observations during natural ROS events and sprinkling experiments, as well as simulations of historical ROS events with the physics-based snow cover model SNOWPACK, lay the foundation for results presented in this thesis. The experimental work was a valuable way to gain hands-on experience of snow cover processes during ROS and collect data for model development and verification. The simulations of more than 1000 historical ROS events at station locations and 191 catchment-scale simulations, increased understanding of runoff formation processes for a variety of meteorological and snowpack conditions. Meteorological forcing and initial snowpack properties were found to determine the temporal dynamics, intensities, and cumulative amount of snowpack runoff. Processes within the snowpack were found to modulate meteorological forcing such that runoff intensities were attenuated for intense and short rain events, but amplified for longer rain events. Although rainfall generally dominated snowpack runoff, individual events did have a significant snowmelt contribution. The analysis of spatially distributed snowpack simulations allowed identification of conditions leading to excessive snowpack runoff for whole catchments. These conditions include: a large snow-covered fraction, spatially homogeneous snowpack properties, prolonged rainfall events, and a strong rise in air temperature over the course of the event. A combination of these factors increases the probability of snowpack runoff occurring synchronously within the catchment, which in turn favours higher overall runoff rates. For both individual station locations and entire catchments, events with excessive snowpack runoff were more common during autumn and late spring, whereas winter snowpack usually retained part of the rainfall. Lysimeter measurements during sprinkling experiments on cold and dry snowpack could not be reproduced in a satisfactory way using the two present water transport models in SNOWPACK. This was attributed to the formation of preferential flowpaths. To address this problem, a dual-domain water transport scheme accounting for preferential flow was implemented in SNOWPACK. The presented approach was validated using an extensive dataset, comprised of meteorological and snowpack measurements as well as snow lysimeter runoff data for more than 100 ROS events. Simulations of ROS on different initial snow cover conditions revealed that the new model was superior to existing approaches for conditions where field experiments found preferential flow to be prevalent. The research presented in this thesis uses systematic analyses to identify meteorological and snowpack conditions which augment snowpack runoff formation during ROS. It further highlights the importance of correct meteorological forecasting and using detailed snowpack modelling for assessing the flooding potential of ROS in snow-affected catchments.