Thanks for your table, and it is good for research people.

Well, the answer to your question isn't so straightforward. To start, Pressure Temperature and water vapor (and winds) are assimilated state variables. The analysis should provide fields that replicate the observations fairly well. Indeed, there tends to be low variations from one reanalysis to the next in the state fields. This is true for upper troposphere extratropics, but the uncertainty increases closer to the surface and closer to the tropics. Also, water vapor tends to be more uncertain than temperature. One of your statements troubles me: "reanalysis values are as good as observations". Reanalyses are not observations, even for the state variables, it is not valid. Only observations are observations, and reanalyses are a processing of the data. For example, you say "as good as observations", to which I would ask, as good as *which* observations? Reanalyses assimilate sondes, aircraft and radiances in the upper troposphere for temperature, all have a role in the analyzed temperature, and can have differences from any one of the assimilated observations. To assess the uncertainty of a reanalysis, there are some things that can be done. Intercomparisons among reanalyses can define a variance and range of results. Reanalyses may provide the analysis increments, the amount that the state variables are affected by observations (over a long period, quantifies the model bias) or some may provide the innovations (sometimes called feedback files) which provide the forecast and analysis error compared to each observation assimilated. Of course, validation against independent observations is always a good exercise for any reanalysis. I hope that helps.

which one of the new generation (ERA-Interim, MERRA, CFSR) reanalysis are more reliable for tropospheric water vapour studies ?