Digital elevation models (DEMs), represent the three-dimensional terrain and are the basic input for numerical snow avalanche dynamics simulations. DEMs can be acquired using topographic maps or remote-sensing technologies, such as photogrammetry or lidar. Depending on the acquisition technique, different spatial resolutions and qualities are achieved. However, there is a lack of studies that investigate the sensitivity of snow avalanche simulation algorithms to the quality and resolution of DEMs. Here, we perform calculations using the numerical avalance dynamics model RAMMS, varying the quality and spatial resolution of the underlying DEMs, while holding the simulation parameters constant. We study both channelized and open-terrain avalanche tracks with variable roughness. To quantify the variance of these simulations, we use well-documented large-scale avalanche events from Davos, Switzerland (winter 2007/08), and from our large-scale avalanche test site, Vallée de la Sionne (winter 2005/06).We find that the DEM resolution and quality is critical for modeled flow paths, run-out distances, deposits, velocities and impact pressures. Although a spatial resolution of ∼25m is sufficient for large-scale avalanche modeling, the DEM datasets must be checked carefully for anomalies and artifacts before using them for dynamics calculations.