This paper describes the regionalization of flow duration curves for rivers located in the Calha Norte and Xingu hydrologic regions, in the Brazilian state of Pará. Recorded data at 25 stream gauge stations located in the regions were used for model fitting. The flow duration curves were calibrated by using of 5 regression models (power, exponential, logarithmic, quadratic, and cubic). The cubic model was found to be the best-fit to the flow duration curves of the Calha Norte. The exponential model was found to be the best-fit to the flow duration curves of the Xingu. The spatial variation of each model parameter was studied in order to explain its corresponding variation in terms of the drainage area, mean annual precipitation, river length and slope. The models were validated to estimating the flow duration curves at two chosen watersheds of each region, with good agreement of the simulated and observed plotting flows. In mathematical words, the good fitting was represented by the relative mean square error and the Nash-Sutcliffe coefficient calculated for the cubic model (Calha Norte) and for the exponential model (Xingu) . The good performance of the models suggests that they can be used to simulate the flow duration curves for the considered regions.
Hydrological modeling; flows simulation; multiple regression; Amazon
