Extensive amounts of water erosion combined with the magnitude of on-site and off-site impacts have led to considerable reduction in soil fertility and increased production costs in agriculture. Digital elevation models are a useful and established tool to better understand surface processes and to quantify damages. Ongoing developments in camera-based 3D reconstruction allow non-expert users to produce high-quality elevation models. Multitemporal surveys allow geomorphic change detection serving as a valuable tool for geomorphologists and erosion researchers. Despite various benefits, these recent developments are accompanied by novel challenges in the form of small-scale nonerosive surface processes (e.g., swelling/shrinking, consolidation/compaction, and aggregate breakdown). If neglected, these alterations can sum up to distinct errors in soil loss quantification. However, as yet, there have been no studies discussing their influence on surface data sets in erosion research. We present an across-scales approach of visualising and quantifying detailed soil physical changes and their influence on surface morphology at submillimetre resolution. The data sets were acquired at field scale from an unmanned airborne vehicle, at plot scale during an artificial rainfall simulation on an agricultural site, and at microplot scale in a laboratory set-up. Resulting high-precision data sets enable the desired insights in soil movements and surface alterations without preceding particle transport and soil loss. Surface changes sum up to 260 m³/ha on an Andalusian test site and could be falsely interpreted as soil loss (of 286 t/ha). These results recommend cautious interpretation of high-resolution topography in soil erosion research.