Advancements in elevation sampling platforms allow for the creation of fine-resolution DEMs that faithfully capture microtopography. However, fine-resolution DEMs are severely impacted by small-scale surface roughness, which is commonly removed during a DEM generalization pre-processing step. This research sought to investigate and develop a DEM generalization technique that removes surface roughness while preserving microtopography. Specifically, preserving microtopographic drainage features such as gullies, culverts, and tributaries are particularly important for hydrological flow-path modelling applications. Two feature-preserving smoothing techniques were developed and compared against common low-pass filter-based smoothing techniques to evaluate smoothing performance and processing efficiency. Using scale signatures, the drainage preserving smoothing (DPS) algorithm effectively removed surface roughness and accurately preserved microtopographic drainage features demonstrated through lower RMSD and LE90 (0.0444 and 0.0582 for Rondeau Bay) values compared to low-pass filtering techniques. Although DPS has an associated performance penalty, feature-preserving techniques documented higher quality smoothing and are practical DEM generalization techniques.