A statistical method based on a simple uniform flow depth–discharge model for a two-stage main channel-floodplain river section has been developed and tested for data from four hydrometric gauging stations. The depth–discharge rating curve fitting procedure is formulated as a piecewise regression problem. A simulated annealing algorithm is used to obtain the least-squares rating curve parameters, including the main channel-floodplain change-point. Uncertainty of all parameter and discharge estimates is approximated by bootstrap techniques. The application of the methodology to field data showed an accurate agreement between the measured and the estimated depth–discharge relationships. A good agreement between the observed and estimated main channel-floodplain change-points was also achieved. Some difficulties, caused by factors such as unequal measurement error variance, infinite parameter estimates and the presence of more than two depth–discharge change points, were apparent; these are discussed.