Análise de incerteza no escoamento simulado com o modelo distribuído NAVMO.

Abstract

The main purpose of hydrologic modelling is the simulation of the water cycle processes. These processes are complex owing to their variability, particularly at large river basins. Distributed models which may consider these variabilities of the system have been designed recently. However, even with such models the simulations contain uncertainties due to the parameterization, poor quality data, model structure and model user. In this sense, this research intend to analyze the uncertainties in the simulated runoffs at various catchments in the upper Paraíba river basin, with the distributed model NAVMO considering different basin divisions and databases, an aspect not yet investigated regionally. The analyses were carried out in relation to previous work with the model, calibrated for the region divided into 61 sub-basins and used to simulate long-time series of runoff with other basin divisions (10, 21, 31 and 40) and four databases consisting of 16, 18, 35 and 80 rainfall stations. In addition, the sensitive parameters of the model were calibrated with the region divided into 31 sub-basins and used to simulate long-term time series of runoff with other basin divisions (10, 21, 40 and 61), which served as the basis for investigating the uncertainties in the simulations, and for comparisons. The investigations were carried out by analyzing the statistical properties of the simulated series such as: the sum of annual simulated values, the double mass curve, the coefficient of determination, the sum of squared differences, the parametric t-test and F-test, the Wilcoxon test and fitting typical distributions of probability (e.g., Gama, Pearson III, Normal, Gumbel and GEV). The results showed that the simulations were sensitive to the basin divisions, model parameterization and rainfall data. In addition, the uncertainties decreased as the number of basin divisions and quality of the rainfall data increased. Moreover, the results were affected by the model structure in relation to the drainage system and rainfall database, which vary depending on the division of the region.