Abstract
This research had as its objective evaluating if Nakagami distribution achieves in precipitation data analysis, by considering monthly data series over several years, in an attempt to select a useful distribution that helps in the management and administration of those activities that depends on Southern Brazil precipitation index. For this purpose Nakagami distribution was compared with five alternatives: Weibull, Gamma, Log-Normal, Log-Logistic e Inverse-Gaussian. In the analysis, data series belonging to 33 weather stations between January 1970 and December 2014 were used, amounting 396 data series (33 weather × stations 12 months). In order to choose the best distribution were used the Akaike Information, the Kolmogorov-Smirnov, then Anderson-Darling and the Cramér-von Mises criterions. According to these criterions, it can be found that Nakagami and Weibull distributions were selected the greatest number of times (Nakagami: 146 times and Weibull: 100 times). Even though Nakagami distribution is not too used in rain precipitation data description, we recommend its application in the description of monthly precipitation behavior as an alternative of distributions that have been traditionally applied.
Keywords:
climate variation; distributions adjustment; maximum likelihood; Nakagami distribution; precipitation data