Streamflow discharge is usually determined indirectly from measurements of the river stage at gaging stations and through the use of stage-discharge relationships (rating curves). However, in alluvial streams, stage-discharge relationships change continually and, sometimes, quite markedly. Such changes may be caused by major floods, seasonal variations, or long-term secular trends associated with changes in the river channel. Consequently, reliable estimates of discharge using rating curves are not possible unless frequent direct measurements of discharge are made. Such measurements involve appreciable costs, and it is important to evaluate their contribution in increasing the accuracy of estimation of quantities of interest such as mean daily, monthly or annual flow. A methodology for the evaluation of the efficiency of data-collection strategies for alluvial rivers is developed and applied to stations on the Missouri River, U.S.A. A flexible and expedient model describing the variability of discharges and shifts in the stage-discharge relationship is developed. Procedures for the estimation of parameters and the validation of the model using actual data are presented. The calibrated and validated model is then employed in simulations to evaluate the effect of sampling strategies (such as frequency and accuracy of discharge measurements) on the accuracy of estimated mean daily, monthly and annual flow. Curves relating the cost of sampling to the achieved accuracy can be generated, and the optimization of sampling strategies given accuracy or budget objectives or constraints can be achieved. ?? 1984.