Abstract: Thousands of meteorological observations collected from the meteorological services around the world and satellites are used for operating computer codes for Numerical Weather Prediction (NWP) employing different techniques of data assimilation. The Center for Weather Forecast and Climatic Studies of the National Institute for Space Research (CPTEC-INPE) is implementing the global analysis system called PSAS (Phisycal-space Statistical Analysis System) (GPSAS) to generate initial conditions for its global model CPTEC/COLA. The data assimilation process requests measures of the real state of the atmosphere, these measurements are the meteorological observations received from Global Telecommunications System (GTS) / National Institute of Meteorology (INMET), from satellite data, and alternatively from the National Environmental Satellite Data and Information Service (NESDIS). The analysis system gives an indispensable adjust of the model state of weather forecast in agreement with the observations, once there are not observations for all the points of the domain of the models. These observations are processed using standard codes from the World Meteorologial Organization (WMO), and they are prepared in agreement with the requirements of the assimilation system. Aditionally, several developments are made for pre-processing the observational data for the Global PSAS. In this paper methods used for selection and adaptation of obsrvations will be presented, as well as a general view of the space distribution of observations received and used in CPTEC.