GPS can be used to determine the integrated precipitable water vapor (IPWV) on temporal and spatial scales that before GPS were unattainable. We investigate the spatial coherence of IPWV estimates for a trial 5-day period using estimates acquired from the Swedish permanent GPS network. We perform a modified Empirical Orthogonal Function analysis and find that over 90% of the water vapor variability is explained using the first temporal eigenvector only. Using the first eigenvectors, we calculate residual time series. We find that the error spectrum of the IPWV estimates contains no observable white noise component, possibly as a result of the stochastic filtering technique employed to obtain the estimates. These results indicate that GPS estimates of IPWV could be extremely useful for following the spatial progress of airmasses, but more work is required to be able to use these estimates for studying the detailed dynamics of turbulence in the atmosphere.