General circulation model (GCM) cloud parameterizations generally include an assumption about the sub-grid scale probability distribution function (PDF) of total water or a related variable. The choice of the PDF and its parameters is usually based on the behavior of the GCM radiative fluxes because observational estimates are not available. In the present study AIRS monthly mean cloud amount and relative humidity fields are used to compute a diagnostic quantity which is shown to act as a proxy for a the second moment of a GCMs total water PDF. In addition, the dependence of the GCM PDF second moment on horizontal grid resolution is analyzed here using results from a high resolution global model simulation.

The AIRS-derived "RH01 diagnostic" is generally larger near the surface than aloft, indicating a narrower PDF near the surface, and varies most markedly with the type of underlying surface. The RH01 profiles differ somewhat among geographic regions, the tropical profiles being drier aloft and wetter near the surface than extratropical profiles. High resolution model results showed that the vertical structure of profiles of the GCM PDF second moment is unchanged as grid resolution increases from 200 km to 100 km to 50 km, and that the profiles shift towards higher values with increasing resolution.

Several GEOS-5 GCM simulations of 20 years duration were performed with several choices of the PDF second moment profile. The resulting cloud and humidity fields were shown to be quite sensitive to the prescribed profile, and the use of a profile based on the AIRS derived proxy results in improvements relative to observational estimates. The AIRS-guided total water PDF profiles, including their dependence on underlying surface type and on horizontal resolution, have been implemented in the GEOS-5 GCM.