Fixed exact restart when FROZEN_SOIL=TRUE #509
Conversation
for both classic and image driver (This is a temporary solution to ensure exact restart)
- Finished separating `estimate_layer_ice_content`
- Moved calculation of `cell.layer.ice` from
`compute_derived_state_vars` to `generate_default_state`
for clarity purpose; does not affect results
with exact restart)
to `generate_default_state` when QUICK_FLUX=TRUE - this resolves the exact restart issue when FROZEN_SOIL=TRUE and FULL_ENERGY=TRUE for QUICK_FLUX=TRUE
…image_driver_EB_FS
|
A few build warnings that should be addressed: |
|
@jhamman Yes thanks! I'll update those. |
|
@jhamman I've cleaned up some unused variables. |
|
@yixinmao : Please be specific on what was fixed and what still remains for fixing? When you say "some" of the variable that is unclear. |
| if (ErrorFlag == ERROR) { | ||
| log_err("Error in " | ||
| "estimate_layer_ice_content_quick_flux"); | ||
| "estimate_layer_temperature_quick_flux"); |
There was a problem hiding this comment.
Change to:
log_err("Error in estimate_layer_temperature_quick_flux");unless this is an uncrustify result in which case just leave it.
There was a problem hiding this comment.
@yixinmao, what's going on here?
There was a problem hiding this comment.
oh I see. Go ahead and leave as is.
|
@bartnijssen @jhamman : To clarify: I've cleaned up the warnings in response to Joe's comment. |
| Lsum[lidx], | ||
| Zsum_node[min_nidx], | ||
| Zsum_node[min_nidx + | ||
| 1], |
There was a problem hiding this comment.
put the 1 on the same line
|
@yixinmao - ping me for review after you've merged yixinmao#1. |
…EB_FS remove hard-coded, statically allocated array sizes for tmpT and tmpZ
…FS' into restart/image_driver_EB_FS Conflicts: vic/vic_run/include/vic_run.h
|
@yixinmao - we just need an additional comment to the release notes. Just explain this change in the existing exact restarts bug fix section and reference the relevant github issues. |
| size_t frost_area; | ||
| size_t tmpTshape[] = { | ||
| options.Nlayer, options.Nnode, | ||
| options.Nfrost + 1 |
There was a problem hiding this comment.
Cool - I did not know that that actually worked!
This PR fixed the inexact restart when
FROZEN_SOIL=TRUE. Specifically:estimate_layer_ice_contentorestimate_layer_ice_content_quick_flux). Now, the calculation for layer ice content and layer temperatures are done in separated functionscompute_derived_state_varstogenerate_default_state, because layer ice content only needs to be initialized when there is no initial state file; when there is an initial state file, layer ice content should be read from the file. The calculation for layer temperature remains incompute_derived_state_vars, since it is a prognostic state variable.