This paper describes a project combining field studies and analyses directed at providing an assessment of the accuracy of remotely sensed methods for determining river characteristics such as velocity and discharge. In particular, we describe a remote sensing method for surface velocities using mid-wave thermal camera videography combined with image analysis. One of the critical problems in this work is determining a method for relating remotely measured water-surface velocities to vertically averaged velocities through a velocity index. We explore three similarity profiles that allow a relationship between surface and vertically averaged velocity to be found either using empirical results or simple roughness-to-depth ratios. To test the approaches we compare them in a situation where vertical structure is known over most of the flow depth through ADCP measurements. By determining best-fit profiles through the ADCP profiles, average values of the velocity index are found for the cross-sections where measurement were made. By comparing these to the predicted velocity indices from the three similarity profiles, we find that, although the differences between the various similarity profiles are substantial, they are smaller than differences associated with local nonuniformity and nonhydrostatic flow. Nevertheless, the velocity indices are accurate to about +/-5%, meaning that remotely sensed vertically averaged velocities can be computed to well within the current accuracy standard for such values when used for river gaging.