Métodos de interpolação espacial para o mapeamento da precipitação pluvial

Abstract

The mapping of weather elements, especially rainfall, needs constant studies to improve the performance of interpolators, and to obtain unbiased maps. This work aimed to evaluate the performance of some spatial interpolators for mean monthly rainfall, mean rainfall during the dry season and mean annual rainfall in the Minas Gerais State. For that, the ordinary kriging was evaluated and compared, after the semi-variogram modeling, co-kriging (introducing the altitude above sea level as a secondary variable), statistical modeling in which the mean precipitation can be estimated from geographical coordinates, and inverse square distance. In this study, data sets were evaluated from 232 pluviometric stations, in the Minas Gerais area, to apply each one of the interpolators mentioned and 70 stations to evaluate the performance that was conducted on the basis of absolute mean error. In addition, a digital elevation model, with a resolution of 270 m, was applied. The interpolators have shown good performance, with mean errors varying from 12.84 to 19.96%, the co-kriging method presenting a smaller absolute mean error in 50% of the situations evaluated.