Probabilidades de duração de seca empregando a Teoria de Runs

Abstract

The univariate and bivariate structural analysis was made considering monthly an yearly time intervals of four runoffs series and two rainfalls series. A log normal three distribution was adjusted to each independent component of hydrologic series available where the time interval was considered to be a month. In case of yearly time intervals a normal distribution was adjusted. Probability distributions of longest negative runlength in given series of length N = 300, 600 and 1200 (monthly case) and N = 25, 50 and 100 (yearly case) and probability distributions of the run-lengths in infinite series for truncation levels q (q = .3, .4 and .5) that correspond to the demand were determined, considering the univariate and the bivariate case. The distributions were evaluated by the experimental method of Monte Carlo with univariate and bivariate series generation. Comparisons between experimental and exact methods and between experimental and aproximate methods were made as well as an analysis of the representativity of the critical historical situation.