Fix vertical indices of c1h and c2h in ideal case init of ph_1

[traupach]

TYPE: bug fix

KEYWORDS: ideal case, initialization

SOURCE: Tim Raupach (UNSW CCRC)

DESCRIPTION OF CHANGES:
Problem:
In ideal case initialization routines, incorrect indices were used in the calculation of ph_1 after a perturbation bubble was added.

Impact:
The calculation of perturbation geopotential ph_1 (perturbation geopotential) and Z_BASE (idealized base state height)
were affected.

1. If hybrid coordinate option was selected, (hybrid_opt = 2), the entire ph_1 initialization is incorrect due to the offset
   in the indexing of c1h and c2h (half levels vs full levels).
2. If hybrid_opt = 0, since the 1d variables are initialized kms:kme, there is no impact.

Solution:
The vertical indices were corrected in the calculations of ph_1 in the idealized initialization routine.

ISSUE:

LIST OF MODIFIED FILES:
M dyn_em/module_initialize_ideal.F

TESTS CONDUCTED:

1. The jenkins testing is all pass.

RELEASE NOTE: In the ideal case initialization routines, incorrect indices were used in the calculation of ph_1 after a perturbation bubble was added. For the idealized cases that do not use the hybrid vertical coordinate by default, there is no impact at all. The ideal cases with a hill had the correct hybrid formulation.

[dudhia]

I agree with this change.

[dudhia]

My understanding is that this won't affect results unless the hybrid option is used which is off by default for these cases. Can the PR proposer confirm this and add some text in the PR on the impact? Thanks.

[traupach]

My understanding is that this won't affect results unless the hybrid option is used which is off by default for these cases. Can the PR proposer confirm this and add some text in the PR on the impact? Thanks.

Yes, hybrid_opt = 0 for these cases by default. When the hybrid option is off, only the top-most model level is affected by the incorrect indices. For vertical indices 1 to kde-1, c1h is set to 1 and c2h is set to 0 on lines 875 and 876 of module_initialize_ideal.F. However ph_1 is calculated for vertical indices k=2 to k=kte, and with the incorrect vertical indices c1h(kte) and c2h(kte) are used in the calculation of ph_1 for the top-most level (since kte == kde). c1h(kte) and c2h(kte) are unmodified by module_initialize_ideal.F and are both set to 0 (or possibly uninitialized?).

I've added a note about the impact in the PR.

[dudhia]

Thanks fro including the impact in the PR message. Looks fine to me.

[davegill]

@traupach
Tim,
What is the status of the jenkins testing? Did you receive an email? Can you post the text of that email in one of these comment boxes?

[davegill]

Please find result of the WRF regression test cases in the attachment. This build is for Commit ID: 17f29b2, requested by: traupach for PR: #1280. For any query please send e-mail to David Gill.

Test Type              | Expected  | Received |  Failed
= = = = = = = = = = = = = = = = = = = = = = = =  = = = =
Number of Tests        : 19           18
Number of Builds       : 48           46
Number of Simulations  : 166           164        0
Number of Comparisons  : 105           104        0

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

[dudhia]

I agree with this change.

[davegill]

@traupach
Tim,
Thanks for the contribution and finding the bugs! I apologize for the delay, but most of that was due to our broken testing system. We welcome you to continue your contributions back to the model for the benefit of the whole community.

You are now officially a WRF contributor.