Makes mp_zero_out affect only the 'moist' array. New separate flags for scalar/chem/tracer

[MicroTed]

TYPE: bug fix

KEYWORDS: mp_zero_out

SOURCE: Ted Mansell (NOAA)

DESCRIPTION OF CHANGES:

This fixes the side-effect of mp_zero_out being applied not only to the moist array, but also to the scalar/chem/tracer arrays using the same threshold. This behavior would be unexpected from the documentation (readme) which indicated that only mixing ratios were affected. This PR restricts mp_zero_out to the moist array and adds a separate mp_zero_out_all flag to apply it to all the arrays in the off chance that somebody needs to replicate the previous behavior.

ISSUE: Addresses #2007

LIST OF MODIFIED FILES:
Registry/Registry.EM_COMMON
dyn_em/solve_em.F
run/README.namelist
wrftladj/solve_em_ad.F
wrftladj/solve_em_tl.F

TESTS CONDUCTED:
The Jenkins tests have passed.

RELEASE NOTE: The behavior of the mp_zero_out flag was changed affect only the 'moist' array, whereas previously it also caused the scalar/chem/tracer arrays to also be set to zero for values below threshold. Now there is a separate flag (mp_zero_out_all) if one wishes to reproduce the old behavior.

[dudhia]

Thanks. I would advocate eliminating the other new zero_out options completely. The existing routine has too many moisture references and even uses p_qv. Hard to anticipate anyone needing this for other 4d arrays, as I doubt it is used much anyway.

[MicroTed]

@dudhia I don't disagree, as I would not use any of them. But if you google for mp_zero_out, there are recommended namelists that have it set > 0. So I would not be totally shocked if somebody would want to reproduce the original behavior for comparisons. I suppose a single flag and threshold for the chem/scalar/tracer would be sufficient for that, though, and 4 flags is perhaps overkill ...

[dudhia]

I understand your caution about backward compatibility. Another alternative is to leave the old calls in there but commented out in case someone wants to reactivate them easily.

[MicroTed]

Or maybe just a 'mp_zero_out_all' flag to turn it on for all the arrays? (Basically the same as commenting out the extra calls)

[dudhia]

I like that idea.

[weiwangncar]

@MicroTed The regression tests are failing for DA/4DVAR compilation. When the changes are finalized, some of the registry changes should be added to registry.wrfplus and/or registry.var.

[MicroTed]

@weiwangncar Thanks for noting that. I tend to forget about var.
@dudhia I'll make that change to a single extra flag -- much simpler

[weiwangncar]

The regression test results:

Test Type              | Expected  | Received |  Failed
= = = = = = = = = = = = = = = = = = = = = = = =  = = = =
Number of Tests        : 23           24
Number of Builds       : 60           57
Number of Simulations  : 158           150        0
Number of Comparisons  : 95           86        0

Failed Simulations are: 
None
Which comparisons are not bit-for-bit: 
None