The numerical weather models are essential tools for predicting short and long term, allowing the prediction of weather conditions several days in advance. The knowledge of models performance and the systematic errors associated with them is extremely important because it allows to evaluate the ability to capture the physical processes of the atmosphere. In order to improve the quality of weather forecast in South America, available at Center for Weather Forecast and Climate Studies (CPTEC), this study evaluated the forecasts of precipitation and mean sea level pressure for the period up to 120 hours, using the mean error (ME) and the root mean squared error (RMSE) from December 2007 to February 2008. The used model was the ETA (40 km), with two separate entries of data, the analysis of the Physical-Space Statistical Analysis System (PSAS) (ETA-I) and the National Centers for Environmental Predictions (NCEP) (ETA-II). The results showed, for both tests, a trend of overestimation (positive values of average error) of precipitation on the Northern Region of Brazil, mainly for the 24 hours forecast. Considering the mean sea level pressure (MSLP), it was clearly seen that the ETA-I, whose values of pressure are very close to observed, provides better results compared to the ETA-II, especially during the first hours of integration.
models; precipitation; evaluations
