Desenvolvimento de um modelo chuva-vazão via tank model e calibração automática

Abstract

As rainfall affects directly estimations of runoff, this study focused on runoff analysis resulting from the application of a rainfall-runoff model based on tanks - Tank Model. To this end two historical series of precipitation on Manoel Alves River Basin, State of Tocantins, Brazil, have been used. The model parameters were obtained by automatic calibration by using two different optimization methods (nonlinear programming and genetic algorithms) as well as two different intervals for time series data (monthly and daily) in order to compare the results. The results have shown that the simulated data for monthly resolutions were more consistent than those ones obtained for daily data when comparing the observed runoff. The study demonstrated that it is not advisable to use daily data for the calculations of streamflow forecasting by Tank Model. The set of estimated parameters was considered satisfactory, which can enable modeling efficiently. The study could contribute to the generation of reliable synthetic discharge series, which can allow simulations of reservoir behavior for various water use scenarios mainly in areas with poor series data (lack of data or long gaps in data discharge series), which is very common in Brazil.