The goals of this study are 1) to develop an analysis technique for deriving and continuously updating a near real-time, global, 2-hourly data set of atmospheric precipitable water (PW) using existing ground-based Global Positioning System (GPS) measurements of zenith path delay (ZPD), 2) to apply the PW data to study the diurnal variations in PW over the globe, and (3) to estimate the diurnal sampling errors in twice-daily radiosonde humidity, and quantify spatial and temporal inhomogeneity and biases in global radiosonde PW data. The advantages of GPS-derived PW, including high temporal sampling resolution, availability under all weather conditions, long term stability, low cost, and large coverage, make the PW dataset unique for long-term climate monitoring. The GPS ZPD data are currently available every two hours at about 360 GPS ground stations as part of the International GPS Service (IGS) products. This number has been increasing and is expected to continue to increase. Surface pressure and weighted-mean atmospheric mean temperature Tm are required for conversion from ZPD to PW. The analysis technique includes (1) obtaining surface pressure from global, 3-hourly surface synoptic observations, and (2) deriving Tm by using either temperature and humidity profiles from two reanalyses (ECMWF/ERA-40 or NCEP/NCAR) or empirical relationships between Tm and the surface temperature (Ts). Preliminary comparisons of Tm derived from ERA-40, radiosonde, and Bevis92’s Tm-Ts relationship suggest that ERA-40 overestimates Tm, whereas Bevis92 underestimates Tm in the Tropics. Extensive comparisons of Tm from ERA-40, NCEP/NCAR reanalysis, radiosondes and local Tm-Ts relationship are underway to evaluate various estimates of Tm and find the best approach for global estimation of Tm. The GPS-derived PW dataset will be validated by comparing with PW data from radiosondes, microwave radiometer, and other research-quality instruments. Diurnal variations of atmospheric PW over the globe will be studied using this dataset along with its implications for errors/biases in global operational radiosonde humidity data.