Estimação de vazões em locais não monitorados a partir do emprego conjunto de regressão funcional múltipla e krigagem topológica

Abstract

Streamflow estimates are necessary for a variety of uses in water resources planning and management, such as the design of hydraulic structures, the granting of water uses, the design of water supply systems, irrigation and navegation, and studies for polluant dilution and ecologic flows. Frequently, such estimates are sought at ungauged locations, which, in turn, would require the identification of prediction models. In view of this problem, a method for estimating streamflow at ungauged locations from regionalized flow duration curves (FDCs) and topological kriging is developed is this study. Regional FDCs are obtained with functional multiple regression, which, as opposed to traditional regression approaches, allows estimating the quantiles of FDC with a single model identified from physical and climatic characteristics of the catchments. Once the regional FDCs are obtained, the nonlinear spatial interpolation termed Top-Kriging is utilized for transferring information regarding the durations from nearby sites, which allows the generation of synthetic daily streamflow series. The proposed approach was applied in the São Francisco River catchment, located in the Brazilian state of Minas Gerais. Overall, the performance of the method was appropriate in those catchments that did not considerably deviate from the average values of the covariates utilized during model calibration. Furthermore, the goodness-of-fit of the synthetic streamflow series and, accordingly, the model ability of reproducing the observed annual daily maxima, annual daily minima, and monthly averages were directly related to the suitable estimation of the regional flow duration curves.