Adding PV and PT as dynamical tracers to E3SM

Ok so PV and PT are non-zero at the end of inidat, which is contained in the EAM code.

However, at the time that prim_init2 is called within the HOMME code, it has been zeroed out (based on log files that show min/max values).

Where is prim_init2 called?

In eam/src/dynamics/se/dyn_comp.F90, it's called in the subroutine dyn_init2. Note that in idealized and adiabatic model configurations, elem%Q is set to zero.

where is dyn_init2 called?

In am/src/control/startup_initialconds.F90, the routine initial_conds calls read_inidat.

summary of how to handle different cases:

in eam/src/dynamics/se/restart_dynamics.F90, dyn_init2 is called. However, this seems to assume that all Q fields are read in from a restart file (reasonable). We don't actually have to change this! It's reasonable to assume that if you are running a restart file, it was initialized using the dynamics tracers.

It's also called in eam/src/dynamics/se/inital.F90

Breadcrumb: where is PV_TRCR output?

Idea: don't have to change it at output time. We can add/subtract value in EAM glue code before advection is called! Provisionally: modify stepon_run3 in eam/src/dynamics/se/stepon.F90. Wrap the call to dyn_run. Current hypothesis: no matter what advection scheme you use, it's handled in calls to prim_run_subcycle

eam/src/physics/cam/constituents.F90 eam/src/physics/cam/cam_diagnostics.F90 these are places where generic outfld calls are defined. The key finding is that output needs to propagate to the state FV physics object, which is what is passed to the outfld call. Current hypothesis is that we should do the compensation just before passing back to physics