Rain on snow (ROS) has the potential to generate severe floods. Thus, precisely predicting the effect of an approaching ROS event on runoff formation is very important. Data analyses from past ROS events have shown that a snowpack experiencing ROS can either release runoff immediately or delay it considerably. This delay is a result of refreeze of liquid water and water transport, which in turn is dependent on snow grain properties but also on the presence of structures such as ice layers or capillary barriers. During sprinkling experiments, preferential flow was found to be a process that critically impacted the timing of snowpack runoff. However, current one-dimensional operational snowpack models are not capable of addressing this phenomenon. For this study, the detailed physics-based snowpack model SNOWPACK is extended with a water transport scheme accounting for preferential flow. The implemented Richards equation solver is modified using a dual-domain approach to simulate water transport under preferential flow conditions. To validate the presented approach, we used an extensive dataset of over 100 ROS events from several locations in the European Alps, comprising meteorological and snowpack measurements as well as snow lysimeter runoff data. The model was tested under a variety of initial snowpack conditions, including cold, ripe, stratified and homogeneous snow. Results show that the model accounting for preferential flow demonstrated an improved overall performance, where in particular the onset of snowpack runoff was captured better. While the improvements were ambiguous for experiments on isothermal wet snow, they were pronounced for experiments on cold snowpacks, where field experiments found preferential flow to be especially prevalent.