The NCAR Community Climate Model, Version 2, uses the spectral transfo
rm method for the underlying dry dynamical fluid flow component, and a
monotonic, semi-Lagrangian transport algorithm for water vapor specif
ic humidity. The reasons for this choice of 2 different approaches for
these different components are reviewed, and the details of the imple
mentation of the transport algorithm are presented. The properties of
the transport scheme are described in the context of a 20-year control
simulation. This simulation is also compared to one using the spectra
l transform method for water vapor transport. Neither numerical approa
ch is perfect. The semi-Lagrangian method requires arbitrary correctio
ns to conserve mass, and the spectral method requires the filling of n
onphysical negative values of specific humidity. The analysis shows th
at the balance of processes that produces the climate of CCM2 is very
sensitive to the discrete water vapor transport algorithm. The spectra
l transport method clearly has shortcomings. The negative filler is a
dominant process in significant regions and leads to an incorrect bala
nce of the moisture budget. The mass conservation adjustment of the se
mi-Lagrangian method seems to be cosmetic and has no identified advers
e effect on the simulation.