Effects of random data errors on the parameter values for a conceptual model

A conceptual model was used to generate synthetic error-free runoff data from precipitation and potential evaporation records. Random errors were then introduced into all three data records. By automatically and objectively fitting the model to different combinations of error-free and error-contaminated records, the effects of the errors on the fitting were studied. The value of the fitting criterion was found to depend largely on the errors in the runoff record. The variations in the final parameter values for fittings to error-contaminated data were shown to be no greater than for the error-free case. The data were then used to find the sensitivity figures for each parameter by a method that is less localized than the standard perturbation technique.