The Use of Space-Temporal Geostatistics in the Prediction of Maximum Air Temperature

Abstract

Stochastic processes of spatio-temporal nature consist of phenomenons that are characterized by spatial and temporal variability. Currently, it is one of the great growing areas with diverse applications in environmental, geographic, biological, epidemiological sciences, among others. Certainly, conventional statistical methods are not adequate to modeling self-correlated structures in space and time. In fact, there are still major challenges regarding the computational implementation of the geostatistical methodology for the analysis of space-time processes, with emphasis on the spacetime package of the R program used in this study. Thus, this work aims to apply the geostatistical methodology of covariance functions in order to infer about the maximum air temperature of the State of Minas Gerais from 1996 to 2016, aiming to contribute with challenges such as heating uncontrolled urbanization, scarcity of natural resources, epidemics and natural disasters. Using the data from 61 meteorological stations, the geostatistical space-time analysis was performed, in which the sum-metric covariance model was the most adequate, considering the criterion of the Mean Squared Error. Thus, it was possible to prepare maps of predictions of maximum air temperatures in the state of Minas Gerais through of ordinary kriging, assuming first order stationarity of the evaluated stochastic process. It can be observed that the models of space-time geostatistics have shown to be efficient in the space-time studies of maximum air temperatures.