2010 12 21, JM

Description of code in solve_em.F that allows computation of decoupled
horizontal and vertical advective tendencies.  This is a feature that
was added John Wong and the WRF_CHEM developers and involves threading
additional arrays through the RK advection routines.  The developers
also worked with the WRF Dynamics Working Group to add placeholders
so that this capability could be used to compute decoupled horizontal
and vertical advective tendencies for fields in other 4D tracer arrays
(e.g. moist, scalar, etc.).  However, these are only placeholders for
the other tracer arrays and users wishing to make this modification to
solve_em should read this file carefully.  THIS IS AN ADVANCED TOPIC.

The logical variable tenddec defined at the top of the solve_em routine
is set within each loops over tracers for the 4D tracer arrays in WRF.
The variable is passed into the RK advection routines and controls whether
or not those routines do the necessary extra computations and store or
accumulate the results into the 3D arrays h_tendency and z_tendency, also
defined at the top of solve_em. Currently, and with the exception of the
loops over chemical species, tenddec is always set to false.  The chemical
loops and the statement to declare h_tendency and z_tendency as fully
3D are guarded by #ifdef WRF_CHEM.  Otherwise, h_tendency and z_tendency
are allocated only minimal storage to save memory.  As long as tenddec is
false, the RK routines will not attempt to access or modify these arrays.

To enable computational of decoupled horizontal and vertical advective
tendencies for other 4D tracer arrays, first modify the declaration
of h_tendency and z_tendency so they are fully dimensioned (as they are
with WRF_CHEM).  Then, in the Registry, add definitions for the fields for
which you wish to have these decoupled tendencies computed to the 4D array
advh_t and advz_t .  Set up an index array similar to the adv_ct_indices
array that is defined in solve_em when WRF_CHEM is defined and then
set up assignments to that array similar to the assignments use to set
adv_ct_indices in the routine setup_ct_indices(), which is defined in
chem/chemics_init.F .  This will cause the rk_update_scalar routine to
update the fields you have specified in the advh_t and advz_t arrays.
Note that you must use your adv_*_indices array as the last index of
the advh_t and advz_t arrays in the arguments to rk_update_scalar (see
the call to rk_update_scalar for the chemistry species as an example).

Additional note: currently advh_t and advz_t are the only 4D arrays
defined in the Registry for this purpose. Therefore, you can only compute
decoupled horizontal and vertical tendencies for only one non-chemistry
4D tracer array.  If you need to do this for more than one, add another
advh_* and advz_* array to the Registry and management them separately.

Only after you have done this should you modify how tenddec is set in
the scalar loop you are working with.  Use the assignement at the top
of the chemistry tracer loop as an example.