The applicability of a portable, commercially available surface velocity radar (SVR) for noncontact stream gauging was evaluated through a series of field-scale experiments carried out in a variety of sites and deployment conditions. Comparisons with various concurrent techniques showed acceptable agreement with velocity profiles, with larger uncertainties close to the banks. In addition to discharge error sources shared with intrusive velocity-area techniques, SVR discharge estimates are affected by flood-induced changes in the bed profile and by the selection of a depth-averaged to surface velocity ratio, or velocity coefficient (a). Cross-sectional averaged velocity coefficients showed smaller fluctuations and closer agreement with theoretical values than those computed on individual verticals, especially in channels with high relative roughness. Our findings confirm that a 0.85 is a valid default value, with a preferred site-specific calibration to avoid underestimation of discharge in very smooth channels (relative roughness < 0.001) and overestimation in very rough channels (relative roughness>0.05). Theoretically derived and site-calibrated values of a also give accurate SVR-based discharge estimates (within 10%) for low and intermediate roughness flows (relative roughness 0.001 to 0.05). Moreover, discharge uncertainty does not exceed 10% even for a limited number of SVR positions along the cross section (particularly advantageous to gauge unsteady flood flows and very large floods), thereby extending the range of validity of rating curves.