Three-dimensional data acquired by terrestrial laser scanning (TLS) provides an accurate representation of Earth's surface, which is commonly used to detect and quantify topographic changes on a small scale. However, in Arctic permafrost regions the tundra vegetation and the micro-topography have significant effects on the surface representation in the captured dataset. The resulting spatial sampling of the ground is never identical between two TLS surveys. Thus, monitoring of heave and subsidence in the context of permafrost processes are challenging. This study evaluates TLS for quantifying small-scale vertical movements in an area located within the continuous permafrost zone, 50 km north-east of Inuvik, Northwest Territories, Canada. We propose a novel filter strategy, which accounts for spatial sampling effects and identifies TLS points suitable for multi-temporal deformation analyses. Further important prerequisites must be met, such as accurate co-registration of the TLS datasets. We found that if the ground surface is captured by more than one TLS scan position, plausible subsidence rates (up to mm-scale) can be derived; compared to e.g. standard raster-based DEM difference maps which contain change rates strongly affected by sampling effects.