In this study, we compare precipitable water vapor (PWV) values from two different observing techniques over the Caribbean Sea, Gulf of Mexico, and United States regions, including ground-based GPS water vapor sensing and COSMIC radio occultation (RO) measurements, as well as global analyses from NCEP and ECMWF models. The results show good overall agreement; however, the PWV values estimated by ground-based GPS receivers tend to have a slight dry bias for low PWV values and a slight wet bias for higher PWV values, when compared with GPS RO measurements and global analyses. This may be a consequence of missing low altitude data from RO in areas where the water vapor is more concentrated, locations of the RO soundings with respect to ground-based stations, time and distance differences of either GPS techniques, or simply the difference of these two measurement techniques.

An application of a student T-test to this 910 matched profiles sample gives a retrieved t-value of 2.35, which is larger than the critical value of the sample, 1.96. This means there is a significant difference between both ground-based and space-based GPS techniques datasets with a 0.01% chance of observing a difference as large as it was observed in other random samples. The dry bias associated with space-based GPS is attributed to the missing low altitude data, where the concentration of water vapor is large. The close agreements between space-based and global analyses are due to the fact that these global analyses assimilate space-based GPS RO data from COSMIC, and the retrieval of water vapor profiles from space-based technique requires the use of global analyses as the first guess. As a result, there is mutual dependence between them.