GPS water vapor-observing systems have several advantages over other moisture sensing systems. Examples include: high measurement accuracy; arbitrary temporal resolution; all weather operability (i.e. they provide data under conditions when other observations fail or provide degraded data); no requirement for calibration; high reliability; and low acquisition and maintenance costs. The principle shortcoming of the system is that it does not provide, in itself, information about the vertical distribution of water vapor in the atmosphere. Improvements in this area must come from the use of GPS and other observations and data assimilation systems. Since data assimilation is essentially a sophisticated interpolation problem, it requires a detailed knowledge of the forecast model errors and observation errors. In this paper, we will discuss the major sources of GPS water vapor retrieval errors, including biases and spatial/temporal correlations, and evaluate how the errors associated with different upper-air observing systems, especially radiosondes, interact within the 20 km Rapid Update Cycle.