Sea ice surface roughness plays a critical role in snow redistribution and accumulation; however, the spatial scale at which its influence is most effective remains poorly constrained by field observations. During the 15th Chinese National Arctic Research Expedition in 2025, synchronous measurements of ice surface elevation and snow depth were conducted using a lightweight Global Navigation Satellite System (GNSS) Real-Time Kinematic (RTK) system. Observations were collected over multi-year sea ice in the central Arctic north of 85° N, encompassing both level ice and deformed ice regimes. Multi-scale moving-window analysis revealed a strong scale dependence in the statistical relationship between sea ice surface roughness and mean snow depth. The ensemble-mean correlation coefficient reached a maximum at a window radius of approximately 7–10 m (r ≈ 0.56), indicating that this scale dominates the coupling between surface roughness and snow redistribution. At this characteristic scale, rougher ice surfaces generally corresponded to greater snow depths, reflecting the enhanced trapping and retention of wind-blown snow by complex surface topography during wind-driven transport. These findings provide observational constraints for parameterizing snow redistribution processes on sea ice and offer guidance for selecting appropriate spatial resolutions in remote-sensing products aimed at resolving roughness-controlled snow processes.