Can we trust water vapor observations from soundings?

There has been a remarkable difference between the integrated precipitable water (IPW) measured by NWS soundings taken at the KSLC airport and that measured by ground-based GPS receivers the past few days.  As shown below, IPW measured by NWS soundings since Sep 29 has consistently been lower than that measured by the ground-based GPS receivers.

Granted, it has been very dry, but a large sounding-based dry bias remained at 00 UTC Oct 3 when the ground-based GPS measured about 1 cm and the NWS sounding came it at around 0.65 cm.  

I personally put more stock in the ground-based GPS measurements.  There have been several reports of serious deficiencies in NWS sounding observations of late.  One has to wonder how this influences the initial analysis of the NWP models, as well as reanalyses like the NARR which likely are not assimilating the GPS data.