The forecast of meteorological parameters at a given location, such as a meteorological station, is still a challenging problem and an open question. The current weather forecasting tools, based on numerical techniques, are not always able to capture local variabilities of the weather. Thus, statistical techniques which take model outputs and local observations into account have been very useful to compensate systematic errors of those models. Therefore, in order to improve the current forecast system, in this work we shall present the preliminary results of two techniques that we implemented at the Instituto Nacional de Meteorologia, in Brazil. One is the well known Model Outputs Statistics, which is based on linear multiple regression methods. Another technique is based on Neural Networks (NN) approaches. Although both techniques are still under development, the results up to now have shown great potential for their use.