Analysis of monthly and annual rainfall variability using linear models in the state of Mato Grosso do Sul, Midwest of Brazil

Abstract

AbstractThis study aims at fitting polynomial models based on latitude, longitude and altitude coordinates to estimate the monthly and annual mean rainfall in Mato Grosso do Sul state, Midwest region of Brazil. Furthermore, its target is to verify whether hydrologically similar regions provide statistical improvement in the regression fit for rainfall estimation. To create the monthly and annual rainfall for 32 rain gauge stations, there were used at least 15 years long data records for analysis with a percentage of gaps at most 10%. Generally, the monthly and annual rainfall models present suitable statistical validation coefficients. The number of predictor variables enhances the performance of the regression method when estimating monthly and annual rainfall. Fitting regressions in hydrologically similar groups through cluster analysis tends to increase regression performance; however, the limited number of rain gauge stations in Brazil makes this technique difficult to apply because the number of parameters of the regression models may be greater than the number of rain gauge stations in the cluster. The first degree polynomial regression proved to be the most adequate to represent the mean monthly and annual rainfall because of the equivalence between observed and predicted values of rainfall and because of the statistical analysis. Fitting polynomial models presents suitable method for practical applications, forming an important tool for environmental management in the Mato Grasso do Sul State.