GPS data (for this study, the delays -- which are used to compute the precipitable water) were received from two different sources: from Yoaz Bar-Sever at JPL and from Chuck Demets at the University of Wisconsin-Madison. This page outlines shows how very nicely the two data sources compare. The data that is plotted here is the delay; Yoaz used station data (at the surface) to compute the atmospheric correction for the delay -- from which the wet delay is inferred; Chuck used climatology. The two big causes of density changes are temperature (lower temperature, higher density) and amount of water vapor (more water vapor, less density).
The precipitable water is an average for a cylindrical volume with circular diameter of approximately 30 kilometers -- this resolution is a bit finer (meaning it covers a smaller area) than the resolution for the GOES precipitable water.
There are many ways to process the data as well; For this study, precipitable water that was computed (by Yoaz Bar-Sever) using GPS data processed with GIPSY (GPS Inferred Positioning SYstem) software developed at JPL. In addition, Wayne Feltz provided a snippet of precipitable water computed using GaMIT software (where GaMIT means GPS at MIT -- the comparison between GPS and GaMIT can be seen here). So far, none of the results here use Bernese GPS processing software (which software was developed at the University of Berne in Switzerland).
Click here for study results.
Click here for more information on GOES precipitable water and things to keep in mind about it.