AbstractThe dynamics of bed configurations and the effect of the bedforms on flow in sand-bed rivers are among the most challenging aspects of fluvial hydraulics to quantify. The big unknown is the resistance coefficient, without an accurate knowledge of which there is not much point in striving for overly precise numerical simulations of water-surface elevations and velocities. Two basic approaches have been used to account for the influence of bedforms on flow resistance. The first method divides the total resistance into two components: one related to impedance generated by the channel boundary without bedforms (grain resistance) and the other related to the opposing force produced by the bedforms (form resistance). The second tactic predicts the total resistance based on the overall flow and sediment parameters. It is the second approach that is followed in this study in which a neural network model is devised based on the evaluation of 941 measurements of reach-averaged flow resistance in small to large sand-bed rivers ranging from shallow to extremely deep. The model considers the influence of the several controlling variables, including water temperature and its effect on fluid viscosity, to make high-quality predictions of a resistance coefficient.