-
Notifications
You must be signed in to change notification settings - Fork 7
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix for CatchCN CLM4.5 fail on SLES15 GNU #898
Conversation
@gmao-jkolassa : We'd like to get feedback the following:
|
@biljanaorescanin Thank you for implementing this fix. I am happy with this solution for now. @gmao-rreichle , to answer your questions:
The fire_m_fac parameter controls the sensitivity of combustibility to the soil moisture conditions. The large value of 10e15 was chosen to address excessive amounts of fires simulated for PFTs 4 and 6. At the time we tested a few different values for fire_m_fac and found that only by limiting the occurrence of fire in PFTs 4 and 6 to very dry conditions were we able to get a somewhat reasonable match to observations.
Because of the above issue and many other problems with the CLM4.5 fire code that we found, we decided to no longer conduct fire simulations with this code and instead moved forward the implementation of Catchment-CN5.1, with its (arguably) improved fire code. That is to say, no one is currently using Catchment-CN4.5 for fire simulations and so I think setting fire_m_fac to 10.e5 to avoid the issues that were encountered during testing is fine.
I agree that there probably is a better way to address this issue. At the time we were discussing to also vary the parameters controlling the sensitivity of the combustibility to the atmospheric humidity and atmospheric temperature instead of just altering the soil moisture sensitivity alone (which led to having to set it to this very large value). But because of all the other issues mentioned above, we never did test this and instead decided to move on to a newer model. For the same reason and given that Catchment-CN5.1 is around the corner, I am not sure whether it makes sense to spend a lot of time on coming up with a better fix. But, I will let @rdkoster chime in on that point. |
For my part, I don't see the value in putting time and effort into CatchmentCN-4.5. We had high hopes in the beginning, but the value of that version never materialized - even the NCAR people threw up their hands in regard to it and moved on. I'm not sure who would ever want to use it. -- Randy
|
Per @gmao-jkolassa's report of older results from the time when the original 10e15 numbers were implemented, this change would impact the science output of CNCLM45 and thus cannot be merged without proper science evaluation. Since NCAR advises that CLM45 should not be used anymore, the plan is to disable CNCLM45 in GEOS going forward by disallowing LSM_CHOICE=3 in the setup of the GCM and the LDAS. This will be done in separate PRs. |
During SLES15 testing we got CatchCN clm 4.5 model run fail due to floating-point exception.
Issue happens during specific calculation where exponent is too large.
Error is in this calculation:
GEOSgcm_GridComp/GEOSagcm_GridComp/GEOSphysics_GridComp/GEOSsurface_GridComp/GEOSland_GridComp/GEOScatchCN_GridComp/GEOScatchCNCLM45_GridComp/CLM45/CNFireMod.F90
Lines 695 to 697 in e2b79bd
Issue: #896
@gmao-jkolassa this fix works and results in round off difference for CatchCN. Maybe there is better way to do it I am not sure.
For GCM this is zero diff trivial since fix is in CLM4.5 module.
fyi: @gmao-rreichle @rdkoster