We present a technique for diagnosing mechanisms controlling the water vapor distribution, given data sets for the velocity, temperature, and humidity. The technique involves defining a large number of tracers, each of which represents air which has last been saturated in a particular region of the atmosphere. The time-mean, zonal-mean tracer fields show the typical pathways that air parcels take between one occurrence of saturation and the next. Because saturation vapor pressure is a function only of temperature, and because mixing ratio is conserved for unsaturated parcels, these tracer fields can be used together with the temperature field to reconstruct the water vapor field. This method clarifies the separate influences of temperature and circulation on the water vapor field.

The technique is applied first to an idealized GCM in which the dynamics are dry and forced using the Held-Suarez thermal relaxation, but the model carries a passive water-like tracer which is emitted at the surface and lost due to large-scale condensation with zero latent heat release and no condensate retained. The technique provides an accurate reconstruction of the simulated water vapor field, and application in this context allows us to assess some of the (relatively small) errors associated with resolution and averaging. The technique is then applied to the NCEP/NCAR Reanalysis for periods during the winters of 1997 and 1998.

In both the idealized model and the Reanalysis, the dry air in the subtropical troposphere is produced primarily by isentropic exchange with the extratropics. While it is still true, as assumed in some recent studies, that most of the water vapor in the subtropical dry zones comes from the tropics, the implication of our results is that, under climate change, humidity changes in the subtropical dry zones could be driven to a large extent by extratropical processes, to the extent that these could change the fraction of subtropical air which experiences extratropical isentropic drying.