Atmospheric reanalyses provide a consistent representation of the three-dimensional distribution of water vapor and its changes over time. However, due to changes in the observing system, the value of reanalyses at characterizing low-frequency moisture trends is still questioned. In addition, model biases in data spare regions and biases and inhomogeneities in the data being assimilated add further uncertainty. This study uses an independent global integrated water vapor (IWV) dataset derived from reprocessed measurements of the International Global Navigation Satellite System (GNSS) Service (IGS) network of ground-based stations to assess two modern global reanalyses (ERA-Interim and MERRA-2). We inter-compared the means, interannual variability, and trends in IWV of the reanalyses for the period 1980-2016, and compared them to the data from 70 GNSS stations for the period 1994-2014. Good consistency is found in the means and variability of both reanalyses. Their agreement with the GNSS data is good, except at a few sites where representativeness errors are suspected. Trends in IWV in ERA-Interim and MERRA-2 are shown to be consistent with GNSS observations and generally well correlated with surface temperature trends (e.g. moistening usually associated with warming), but MERRA-2 presents a more general global moistening trend compared to ERA-Interim. Large inconsistent trends are found in ERA-Interim over Antarctica and most of the southern hemisphere. Discrepancies in both reanalyses are also evidenced over central and northern Africa, as well as over Indonesia, the Indian Ocean, and central Asia. While northern Africa and Antarctica certainly suffer from poor coverage in moisture observations and model biases, the bad results in other regions suggest changes in the global observing system might have a significant impact on trend estimates and to a smaller extent on variability.