COMPARAÇÃO ENTRE MÉTODOS DE INTERPOLAÇÃO ESPACIAL PARA A ESTIMATIVA DA DISTRIBUIÇÃO DE PRECIPITAÇÃO NO CEARÁ-BRASIL

The spatial distribution of precipitation is still largely represented by geostatistical methods of interpolation and in semiarid areas requires recurrent studies due to their temporal variability. So, the present research aimed to compare the performance of five interpolation methods: Inverse of Square Distance (ISD), Kriging with a spherical and an exponential semivariogram model, Natural neighbor and Spline regularized or also denominated of surface of minimum curvature. For that, were used data of annual average rainfall of a period of twenty years (1991 to 2010) of 252 rain gauges. As a form of evaluation and determination of the most appropriate method, the technique of cross-validation was chosen as the criterion of comparison, determining the root mean square error (RMSE), mean deviation (MBE) and coefficient of determination (r²) between the estimated data and the observed data. The results show that the interpolation by exponential Kriging method resulted in a smaller mean square error (164,09), showing that it is the interpolator of better spatial representation for the data set under study, however the method of interpolation by Spiline obtained the highest coefficient of determination (0,5112) which defines a better arrangement between the estimated values and the observed values.