Comparison of Atmospheric Water Vapor in Observational and Model Data Sets

The global water vapor distribution for five observational based data sets and three GCM integrations are compared. The variables considered are the mean and standard deviation values of the precipitable water for the entire atmospheric column and the 500 to 300 hPa layer for January and July. The observationally based sets are the radiosonde data of Ross and Elliott, the ERA and NCEP reanalyses, and the NVAP blend of sonde and satellite data. The three GCM simulations all use the NCAR CCM3 as the atmospheric model. They include: a AMIP type simulation using observed SSTs for the period 1979 to 1993, the NCAR CSM 300 year coupled ocean--atmosphere integration, and a CSM integration with a 1% CO2 increase per year. The observational data exhibit some serious inconsistencies. There are geographical patterns of differences related to interannual variations and national instrument biases. It is clear that the proper characterization of water vapor is somewhat uncertain. Some conclusions about these data appear to be robust even given the discrepancies. The ERA data are too dry especially in the upper levels. The observational data evince much better agreement in the data rich Northern Hemisphere compared to the Southern. Distinct biases are quite pronounced over the Southern Ocean. The mean values and particularly the standard deviations of the three reanalyses are very dependent upon the GCM used as the assimilation vehicle for the analyses. This is made clear by the much enhanced tropical variability in the NCEP/DOE/ AMIP reanalyses compared the initial NCEP/NCAR Reanalysis. The NCAR CCM3 shows consistent evidence of a dry bias. The 1% CO2 experiment shows a very similar pattern of disagreement with the sonde data as the other integrations, once account is taken of the warming trend. No new modes of difference are evident in the 1% CO2 experiment. All the CCM3 runs indicated too much Tropical variability especially in the western Tropical Pacific and Southeast Asia. A EOF analysis of the interannual variations of the zonally averaged precipitable water and the 500 to 300 hPa layer reveals fundamental differences in the structure of the variations. The impact of ENSO and variations of the ITCZ have only a low level of correspondence between the observed data, much less the simulations. It is apparent that an adequate characterization of the climatology of the global water vapor distribution is not yet at hand