ON THE INFLUENCE OF ERROR MODEL IN THE GOOD PERFORMANCE OF THE HYDROLOGICAL MODEL FOR THE RIGHT REASONS

Autores UPV
Año
CONGRESO ON THE INFLUENCE OF ERROR MODEL IN THE GOOD PERFORMANCE OF THE HYDROLOGICAL MODEL FOR THE RIGHT REASONS

Abstract

Hydrological models provide extrapolations or predictions, which are not lacking of uncertainty. One phase of the hydrological implementation process, significantly contributing to that uncertainty, is the calibration phase in which values of the unknown model parameters are tuned by optimizing an objective function. Traditionally, the most commonly used fitting criterion has been the simple least squares (SLS), regardless of the SLS criterion involves strictly assumptions about the probability distribution of the errors. Failure of these assumptions introduces noise into the estimation of the parameters. In the present research it has been carried out an estimation of the parameters of TETIS in an exercise of Bayesian inference, with three error model assumptions. The main motivations for the development of this work are: (1) verify whether and to what extent a distributed conceptual model calibrated with the SLS criterion become a data-driven model, yielding sets of unrealistic parameter values but with outstanding performance, (2) contrast model divergence phenomenon between the inference layouts, and (3) compare the reliability of the uncertainty assessment for each of the error models. The analysis of the results suggests that to achieve the objective of having a calibrated hydrological model, which works well for the right reasons, it is necessary to draw the inference of its parameters using an appropriate error model, and the traditional SLS calibration criterion, is not adequate for this purpose. (ISBN- 978-0-692-28129-1)