Estudo da evapotranspiração de referência para projetos de irrigação: uma abordagem probabilística

Abstract

The variability of the reference evapotranspiration (ET0) has been widely explored in the literature, being a phenomenon of vital importance in several areas, among them, for the adequate sizing of the capacity of irrigation systems. Its use has been recommended in probabilistic terms, and functions that estimate the probability associated with this phenomenon have been applied. Considering the random behavior of reference evapotranspiration in different regions, it is necessary to select appropriate distributions for each regime. To improve the prediction accuracy of ET_0 data, the introduction of superior prediction models such as probability distribution mixing models is a topic of research interest. In order to evaluate which probability distribution model best fits the reference evapotranspiration data from the municipality of Bom Jesus da Lapa - BA, looking for irrigation system projects, ten probability distributions were recorded, seven of which are standard distributions of one component: Beta, Gamma, Log Normal, Normal and Weibull with two parameters; Burr XII with three and four parameters, and three mixing distributions with two components: the BurrXII PDF and Generalized PDF of Extreme Values (BurrGev7), two-component PDF of Weibull (WW5) and Generalized PDF of Extreme Values with Weibull (GevW6) , evaluating the adjustment for different irrigation periods (1, 2, 3, 4, 5, 6, 7, 8, 10, 12 and 15 days.). Sample statistics were used to select the best model to represent the event frequency distribution for the study region. The mixing probability densities functions were the ones that best fit the observed data, in all analyzed periods. The BurrGev7 mixture distribution was the one that presented the best performance in most periods, followed by FDP WW5 and in third place GevW6. Among the one-component standard PDFs, the two-parameter Weibull showed the best performance, however, inferior to the mixture PDFs.