Add ctsm_sci test list that would be used for releases and making sure important resolutions run well #1142
Comments
ekluzek
added
enhancement
billsacks
removed
the
enhancement
|
I've started this in #1150. I just included the list of compsets that have "science_supported" clicked. So only f09 and f19. There are some issues that I think are just due to known problems. So I'm bringing it in with the following fails... SMS_Ld5.f09_g17.IHistClm45BgcCruGs.cheyenne_intel.clm-default Since there are scientists that run on izumi it would be good to have f09 and f19 tests there. We wouldn't add the CAM grids to izumi though. |
|
@ekluzek I know this is a work in progress, but it sounds like the immediate needs have been met. Should we close this issue? |
|
I suppose we could close it as the test list now exists. |
CNFire: btran2 fixes and general cleanup
(1) Call routine to calculate fire's btran2 from CNFireArea; this has a
few advantages:
- It makes the logic of CNFireArea more clear (rather than depending on
a btran2 variable that is calculated from some other part of the code)
- This avoids having the biogeophysics depend on the biogeochemistry
- This lets us avoid doing this btran calc if using no-fire – or other,
future, fire methods that don't need it
Note regarding testing: In the initial step, I kept this calculation
dependent on a saved version of h2osoi_vol to avoid changing
answers; I changed this in the answer changes in step (2), as noted
below.
(2) Answer-changing fixes to CNFire's btran2 calculation and use:
(a) Calculate fire btran2 using updated h2osoi_vol (this is an
answer-changing cleanup step from (1))
(b) TEMPORARY CHANGE (reverted in the cleanup in (3)): Reinitialize
fire btran2 to spval for all patches in each time step, so that
the fire code isn't trying to use btran2 from earlier time steps
for patches that were in the exposed veg filter at one point but
no longer are.
One implication of this is that, if there is currently no
exposed veg on a column, the new code leads to the block of code
that forces fire_m = 0 (because wtlf will be 0). Previously, in
contrast, it looks like fire_m was allowed to be non-zero even
if there is currently no exposed veg, because btran2 and wtlf
were accumulated if a patch ever was exposed in the past.
(c) Limit fire btran2 to be <= 1, rather than letting it be slightly
greater than 1. (Due to a conditional in CNFireArea, these
slightly-greater-tan-1 values were being ignored when computing
btran_col, rather than averaging in a 1 value.)
(3) Non-answer-changing fire code cleanup:
(a) Cleanup of the btran2 fixes, including reverting the TEMPORARY
CHANGE noted in (2b), instead relying on a better mechanism:
just doing the calculations of btran_col and wtlf over the
exposedvegp filter. Also, get rid of the checks for
shr_infnan_isnan(btran2(p)) and btran2(p) <= 1 (allowed by the
other changes in (2) and (3)).
(b) Set btran2 to 0 over non-exposed-veg points: this changes
answers for the BTRAN2 diagnostic field, but nothing else. (This
follows what is done for the standard BTRAN.)
(c) Move calc_fire_root_wetness for CNFireLi2021 into the base type
to avoid future bugs (assuming that the next fire module will
extend the base class but will also want to use this new version
of calc_fire_root_wetness).
(d) Change fire looping structure to be more standard
(4) Remove some very expensive tests from aux_clm, putting some in the
new ctsm_sci test list instead
(5) A bit of other minor cleanup
Issues fixed:
- Resolves #1139 (Decrease expense of ne0ARCTICGRISne30x8
test and C96 tests)
- Resolves #1153 (Fire btran2 is only computed for exposed
veg patches, but is used over all veg patches)
- Resolves #1170 (CNFire code: btran2 should not be skipped
when it's greater than 1)
- Partially addresses #1142 (Add ctsm_sci test list that
would be used for releases and making sure important resolutions run
well)
…rv_term_reporting Carbon starvation termination mortality reporting
We think it would be good to have a "ctsm_sci" test list that would have the list of important resolutions/configurations used for CTSM science. I think this would be f09_g17, g19_g17, and f45. It should also have a NWP configuration. Anything that we list as "scientifically supported" should be in this test list. Probably would all be for production compiler settings, maybe just one test at debug to make sure there aren't problems with it.
This would also have the resolutions that are critical for CAM (so C96 and ARCTIC, CONUS, ARTICGRIS grids). It would just be for the standard compiler on the main production machine (cheyenne_intel right now).
This test list would be run supplementary to the regular aux_clm test that would be used for normal tags. This would be run for releases, or important tags that we want to make sure the science resolutions are running. We could move some of the tests we have in aux_clm to this test list, to save time in the normal test list. This would be thought of as a supplementary test list that's run in addition to the standard test list done for tags. As such it doesn't need to test the same things already done for aux_clm.
This relates to these issues that talk about testing...
#1139
#1141
The text was updated successfully, but these errors were encountered: