Long run restart fix #1098
Long run restart fix #1098
Conversation
|
Update, this set of changes does enable some tests to pass the restart compare test that did not previously (ie ERS 5month f10/f45), however longer ERS tests a la 10 months still do not pass. |
glemieux
added
the
PR status: Not Ready
|
Oh boo. Should we push these changes anyway and start a new line of investigation for the subsequent failure? Just checking whether it is nocomp or full fates mode that fails? Or both? |
|
Can confirm, ERS_Lm11... passes, ERS_Lm12 and 13 do not. |
|
Can you decide those for me @mvdebolskiy ? (Are they both nocomp?) |
|
@rosiealice both nocomp. Also spitfire mode 0. |
… so it doesnt happen an extra time on the restart call (unlikely but possible)
|
All, I've made some more updates. See comments in #1051 regarding spread. I also realized that we had not added the variable that remembers a plants net carbon uptake at the leaf layer, to the restart. This finally adds it (although its a memory hog, but a necessary one right now). |
rgknox
removed
the
PR status: Not Ready
| @@ -1124,25 +1133,35 @@ subroutine define_restart_vars(this, initialize_variables) | |||
|
|
|||
| ! Only register satellite phenology related restart variables if it is turned on! | |||
There was a problem hiding this comment.
Minor cleanup; since we use hlm_use_sp in a bunch of places still, I think moving this up to be included in a general comment would be ideal. Or we can strike it altogether since it's not relevant in this specific place in the code anymore.
There was a problem hiding this comment.
discussed with @glemieux will incorporate into future PR
|
It looks like this fixed #897 and ESCOMP/CTSM#667. This can be seen via #1096 (comment), which is the next pull request to be tested after this was integrated. |
… next (PR #6018) This pull requests updates the ed_update_site call in elmfates_interfacemod to pass a flag for when this procedure is called during restart. This update should be coordinated with NGEET/fates#1098, which addresses the long duration exact restart issue NGEET/fates#1051. Additionally this pull request resolves #5548 by expanding the fates regression test coverage to include more run mode options for fates at a variety of resolutions and runtimes. [non-BFB] for FATES Fixes #5548
This pull requests updates the ed_update_site call in elmfates_interfacemod to pass a flag for when this procedure is called during restart. This update should be coordinated with NGEET/fates#1098, which addresses the long duration exact restart issue NGEET/fates#1051. Additionally this pull request resolves #5548 by expanding the fates regression test coverage to include more run mode options for fates at a variety of resolutions and runtimes. [non-BFB] for FATES Fixes #5548
Description:
We had been calling a routine (canopy_structure()) during the restart read process. The purpose was to facilitate rebuilding the canopy. However, this had an inadvertent effect of changing results from base, because this procedure would not just rebuild, but modify the cohort compositions due to termination and promotion/demotion in the canopy. We realized that we could simply bypass this call, and add two variables to the restart file to achieve b4b restarts in tests.
This is an API change PR, and needs to be synchronized with ESCOMP/CTSM#2199
Fixes #1051
Collaborators:
@mvdebolskiy identified the locus of the problem which enabled its fix, with team contributions and troubleshooting from @mvertens , @glemieux , @rosiealice and @ckoven
Expectation of Answer Changes:
No answer changes aside from a fixed bug.
Checklist
If this is your first time contributing, please read the CONTRIBUTING document.
All checklist items must be checked to enable merging this pull request:
Contributor
Integrator
Documentation
Test Results:
CTSM (or) E3SM (specify which) test hash-tag:
CTSM (or) E3SM (specify which) baseline hash-tag:
FATES baseline hash-tag:
Test Output: