This study deals with the periodic regression analysis of the monthly total rainfall, for eight locations in the State of Paraná, Brazil. Mathematical models which describe the course of this variable observing a period of twelve months were applied. The analysis was developed by means of matrice algebra and the expansion in Fourier series of a function f(t), periodic and defined for intervals. The original data were adjusted for 30-day months and all harmonic components of the functions referring to each of the eight locations were determined. After testing for normality of the harmonic components, an analysis of variance was made, which made it possible to identify the significant harmonic components which were chosen to make up part of the regression equation corresponding to each location. These equations explain 89.56% up to 99.60% of the variation due to months, for the whole set of studied locations. Probability for occurrence of a year in accordance to the model, in percentage, ranges from 14.19% to 68.42%, while probability for occurrence of a year adjusted to mean, ranges from 0.02% to 1.87%.
climatological methods; mathematical models; meteorological data
