Surface image velocimetry techniques such as Large-Scale Particle Image Velocimetry (LSPIV) are increasingly used for non-intrusive flow measurements in rivers; however, converting surface velocities to depth-averaged values remains a key source of uncertainty, particularly under ice-affected conditions. In this study, LSPIV surface velocities and point measurements from an Acoustic Doppler Velocimeter (ADV) were combined to investigate the vertical velocity structure in two adjacent pool sections of the Don River: an open-water pool downstream of a step and an upstream pool partially covered by ice. In the open-water pool, ADV measurements indicated reduced velocities near the bed and a surface-dominated flow structure. Depth-averaged velocity was therefore estimated from LSPIV using a conventional velocity index of 0.80-0.90, consistent with values commonly reported for turbulent open-channel flow. A sensitivity analysis showed that discharge estimates were weakly dependent on the assumed velocity index, with variations of approximately ±6%. ADV data were used qualitatively to support the plausibility of the assumed range rather than to calibrate a specific theoretical velocity profile. In contrast, the ice-covered pool exhibited a markedly different vertical velocity distribution, characterized by a mid-depth velocity maximum and reduced velocities near both the bed and the ice interface. In this case, classical open-channel assumptions were not applicable, and depth-averaged velocity was computed directly by integrating the measured ADV profile. The results demonstrate that surface velocity alone is insufficient to characterize flow under ice cover and that depth-averaged velocity may exceed surface velocity in such conditions.