Calibrações de múltiplos modelos hidrológicos em diferentes passos de tempo: previsão de vazão com dados escassos

Abstract

The scarcity of observed flow data, especially in developing regions, such as the state of Ceará, limits the management of water resources in Ceará, since it relies on hydrological modeling. For this reason, a local study is necessary to evaluate how refined the flow series should be in order to be effective to predict flows at different time scales. In this study, experiments are carried out with watersheds of 21 fluviometric gauge stations in Ceará, using Budyko’s, Fu’s, GR1A, GR2M and GR4J models, in order to investigate the effects of calibrations performed at less refined scales than the conventional ones, based on the NSE coefficient computed at different scales. Experiments using cross-validation are performed, given the previously verified limitations of traditional calibration and validation with fixed periods. The annual models performed well, with an

exception for Budyko's performance at the sediment soil basins. The monthly model performed well when calibrated with observed annual data or even with data from Budyko (except in sedimentary basins). The daily model did not perform well when calibrated on longer time scales, and was rejected already in the early experiments.