Uncertainty Quantification in Discharge Curves of Fluviometric Stations Using Bayesian Inference

The aim of this paper is to present the construction of a probabilistic reliability region around rating curves previously defined for different fluviometric stations, in order to consider the uncertainty in obtaining fluviometric discharge values calculated by using these curves. Bayesian models with MCMC sampling algorithms (Monte Carlo Markov Chain) are constructed and applied to the error probability distributions obtained from the comparison of predicted fluviometric discharge values (resulting from the application of the rating curves) and observed (by conventional methods). For this study, records of three hydrological stations monitored by COPEL, and ten hydrological stations monitored by CEMIG, were analyzed and used. The results showed that the Bayesian approach has proved successful for the proposed objectives, allowing the construction of reliability region, and with it, the evaluation of the uncertainties associated with the use of rating curves.