The Meteosat-9 is a geostationary meteorological satellite of the EUMETSAT. The Spinning Enhanced Visible and Infra Red Imager (SEVIRI) of Meteosat-9 collects data in eleven bands, with nominal ground resolution of 3 km. An additional panchromatic band, with 1 km resolution totalizes 12 channels. One of the simplest applications of these data is the estimation of land surface temperatures (LST) from brightness temperatures. In this work the goal was to estimate the surface temperatures from Meteosat-9 images, corresponding to the data of the automatic ground weather stations of CPTEC/INPE, used to validate these temperatures in Minas Gerais State, Brazil. It was observed that for 18 days data, was only possible to estimate the temperatures to 12 days, because in this period, 6 days were contaminated with clouds. In cases of absence of clouds, for some sites, images could be used for the estimations of LSTs.