Assimilação do padrão de variabilidade das variáveis de estado de um modelo chuva-vazão em esquemas de simulação / previsão hidrológica

Abstract

Conceptual rainfall-runoff models are meant to capture the dynamics of the hydrological cycle through the representation of the main physical processes that occur at the river basin scale. The complexity of the natural process of transformation of rain into flow enforces these models, regardless of their structures, given that the representation of phenomenon is made in a simplified manner. These simplifications associated with the complex spatiotemporal representation of the phenomenon, corroborate to the growing consensus of the technical community about the need to assess the uncertainties associated with the predictions of these models. This study was guided by the perception that given the importance and necessity of using these models, the development of techniques to enable the identification and reduction of the uncertainty associated with its predictions is necessary. For this, as in various studies undertaken in the last decade, we used Bayesian inference techniques, along with Monte Carlo simulation methods for determining the uncertainty associated with estimates of parameters and state variables present in the structure of a conceptual rainfall-runoff model. In the sequence, the pattern of variability of the state variables of this model was analyzed. The hypothesis of this thesis is that the behavior of state variables is able to characterize the errors in the modeled series, since they assimilate all sources of uncertainties associated with the modeling in each time step. The analysis of input-state-output behavior has been developed from the formulation of two procedures for updating the state variables of the rainfall-runoff model evaluated. A detailed impact assessment of state variables update was developed, on specific periods of the series. It was possible to confront some of the methodological approaches and conclusions described in the main current references on the subject. It is believed that the conclusions described herein can be used to support the development of techniques to reduce the overall uncertainty associated with the use of rainfall-runoff models, in both hydrological simulation and forecasting schemes.