Acoustic Doppler velocity meters (ADVMs) provide an alternative to more traditional flow measurement devices and procedures such as flumes, weirs, and stage-rating for open channels. However, the requirements for correct calibration are extensive and complex. In this dissertation, two original independent projects were undertaken to accomplish a similar goal: to improve the accuracy of ADVM without the need for in situ calibration.

For the first, a 3D computational fluid dynamics (CFD) model is used to design a subcritical rapidly varied flow (RVF) contraction that provides a consistent, linear relationship between the upward-looking ADVM sample velocity and the cross-sectional average velocity, in order to improve non-calibrated ADVM accuracy. CFD simulations validated the subcritical contraction in a rectangular and trapezoidal cross section by showing errors within +1.8% and -2.2%. Physical testing of the subcritical contraction coupled with an upward-looking ADVM in a large rectangular flume provides laboratory validation with measurement errors within +/-4% without calibration.

In the second project, an algorithm is developed for real-time estimation of the cross-sectional average velocity of a channel flow using an upward-looking pulsed wave ADVM. The Velocity Contour Weighting Method (VCWM) is applicable to gradually varied flows in prismatic channels (without the RVF contraction). VCWM estimates the average velocity as a weighted average of ADVM bin velocities. Weights are based partly on the flow geometry and partly on the velocity distribution sampled by the ADVM. Collectively, these two factors enable the VCWM to adapt to a relatively wide range of channel geometry and roughness features which fall within the range of those used to develop the algorithm. Expressions for the velocity weights are devised by first applying a validated 3D computation fluid dynamics (CFD) channel flow model to a wide range of flow scenarios including differing channel geometries, discharge rates, depths, and boundary roughness, which is then reduced empirically with the aid of dimensional analysis to obtain velocity weight equations. Application of the method to a large rectangular flume and field testing in trapezoidal irrigation canals shows that the VCWM predicts the average velocity with an error less than +/-5.5%.