Experimental vertical velocity profiles were extracted from 18 series of aDcp discharge measurements. The campaigns were performed in the Saône river at the CNR Saint-Georges gauging station in Lyon, France. Post-processing methods for aDcp data positioning, averaging and depth-integrating have been implemented and applied to the dataset to get 23 mean velocity profiles across the 100 m-wide section. Individual fitting by a logarithmic law gives a satisfactory representation of the profiles and depth-integrated velocities very close to the experimental ones (errors are typically less than 1%). However velocities near the free-surface are overestimated. The 414 non-dimensional profiles can be represented accurately by a unit log profile except near the surface. Indeed the site-specific surface velocity coefficient is estimated to be 0.91 ±0.04 instead of the unit log value (0.88) or even the usual default value (0.85). Fitting wake-law profile coefficients allows to reproduce the near surface velocities better. Such an analysis of site-specific vertical velocity profiles is especially useful for discharge estimate methods based on an incomplete investigation of the flow field (H-aDcp, LS-PIV, radar wave scattering, etc.).