Abstract
Meteorological application of Global Navigation Satellite System (GNSS) data over Brazil has increased significantly in recent years, motivated by the significant amount of investment from research agencies. Several projects have, among their principal objectives, the monitoring of humidity over Brazilian territory. These research projects require integrated water vapor (IWV) values with maximum quality, and, accordingly, appropriate data from the installed meteorological stations, together with the GNSS antennas, have been used. The model that is applied to estimate the water-vapor-weighted mean tropospheric temperature (Tm) is a source of uncertainty in the estimate of IWV values using the ground-based GNSS receivers in Brazil. Two global models and one algorithm for Tm, developed through the use of radiosondes, numerical weather prediction products, and 40-yr ECMWF Re-Analysis (ERA-40), as well as two regional models, were evaluated using a dataset of ~78 000 radiosonde profiles collected at 22 stations in Brazil during a 12-yr period (1999–2010). The regional models (denoted the Brazilian and regional models) were developed with the use of multivariate statistical analysis using ~90 000 radiosonde profiles launched at 12 stations over a 32-yr period (1961–93). The main conclusion is that the Brazilian model and two global models exhibit similar performance if the complete dataset and the entire period are taken into consideration. However, for seasonal and local variations of the Tm values, the Brazilian model was better than the other two models for most stations. The Tm values from ERA-40 present no bias, but their scatter is larger than that in the other models.
