Contribuição à Regionalização de Vazão Mínima de Referência na Amazônia Legal

Abstract

Studies aimed at knowledge flows in ungauged sites, are of great importance in hydrology, because almost always the hydrometric network does not cover all areas of interest. Therefore, it is necessary to use techniques that allow the transfer of local information monitored for ungauged sites. In this context, this paper proposes mathematical models to determine the remaining flow in the Amazon, using the regionalization method of flow. The methodology was based on the minimum flow Q95% obtained from the flow duration curve plotted based on the Weibull equation, in addition to morphoclimatic characteristics of river basins, such as the drainage area (A), average annual precipitation (P) and the river length (L). The data used for regionalization belong to six hydrologically homogeneous groups analyzed in the study. With this information, and in order to define which explanatory variables (A, P and L) would be part of the models to be applied, the multi-colinearity test was performed. Thus, taking the result of the previous test, the mathematical models of regionalization of the minimum reference flow Q95% were constructed through multiple regression models (linear, potential, exponential and logarithmic).In the first analysis, the performance values (mean absolute percentage errors, determination coefficients, adjusted coefficient of determination) were unsatisfactory. For this reason, it was chosen to divide the hydrologically homogeneous groups according to the interval with the same order of magnitude of areas and flows. In this way, the results obtained in the adjustment of the models presented a better performance, which can be verified by the average absolute error below 13.78% in all groups. The models that obtained the best performance were selected for the validation, observing in this way, that the linear model was the one that presented the best results for nine of the groups and subgroups tested, standing out also the potential and logarithmic models, that simulated six of the subgroups each.Thus, from the regional equations that presented good results, both in the adjustment and in the validation, we can obtain the flow rate Q95%, in a basin without flow data according to the homogeneous regions or groups determined in the study, having as data of The physical and climatic characteristics of this basin.