Análises de incertezas na determinação de vazões regularizadas em climas semi-áridos

Abstract

The uncertainties inherent of the random variability of natural discharges are, undoubtedly, the biggest obstacle that the determination of the actual reservoir yield has to face, mostly in semi-arid areas. Such uncertainties cannot be reduced by the elaboration of a more precise mathematical model or by increasing the length of observations. They, simply, exist and must be studied through appropriate methodologies, joining valuable information to the water resources planning. It is very usual simulate a given reservoir using the historical data as being the best estimate for the future discharges. The inconvenience of this analysis is that the historical series is just one from all possible sequences, "raffled" by Nature. The use of Monte Carlo Method, otherwise, provides means for the generation of countless independent sequences of discharge, each one different from the historical one, but with the same statistical properties and as equally probable. Once each sequence is different from the other, several results are obtained from simulation, instead of only one, in case of just the historical series was available, allowing the planner to take his decision based not only on an isolated event, but in a probabilistic analysis of the phenomenon. Although Monte Carlo Method do not extend the historical series, it allows to extract the maximum possible information from the data. This research analyzes, using stochastic simulation, the effects of the reservoir initial storage (V0), coefficient of variation of natural inflows (CVdef), dimensionless capacity factor (f K ) and reliability level (G) on the evaluations and errors of the release dimensionless factor estimates (f M). It also determines the best estimator for the steady state dimensionless discharge (f M*), regarding to unbiasedness and efficiency properties of a good estimator.