-
Notifications
You must be signed in to change notification settings - Fork 18
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
Fixes #1199 - HorizontalFluxRegridder numerical bug. #1202
Conversation
@sdeastham Careful interpreting the above results. Because the input fluxes were constant, the regrid essentially returns
I.e., it just measures how accurate the formulae for the arc lengths, as mathematically |
Errors appear mostly near the poles. I think that is because "longitude" is very sensitive to position in that vicinity and increases error in the trig expressions. |
Thanks @tclune ! So, to clarify, what's being plotted is the error in the new calculation? Something I'm curious about - is the calculation of length identical to the method used in FV3? My thinking is that, if so, then this error shouldn't actually matter (because the same error was present in FV3 when calculating the fluxes). |
No the output fields from a test signal are being output. The variation from "2" is the error (entirely due to finite precisions FP, not to regridding). @bena-nasa has set up a new test that has spatially varying input which will be a better indicator. But harder to interpret. :-) Hopefully I can exercise that next week. |
Ha sorry, that's what I meant - so if
The above wouldn't be true if the regridder calculated the coarse flux as |
Yes. I could try to make it use "d1 + d2". Would not even be a lot of work. It just makes the error more subtle though, I think. |
Wouldn't using |
Do the cool pictures change with this update? I like seeing them. 😄 |
This issue has been automatically marked as stale because it has not had recent activity. If there are no updates within 7 days, it will be closed. You can add the "long term" tag to prevent the Stale bot from closing this issue. |
Pinging @tclune. Should this stay open? |
Nah - we know where to find it if @sdeastham returns to the issue. |
@sdeastham Is this the bug you mentioned in the AIST meeting today? |
It is! I couldn't recall what the conclusion was... |
I've restored the That said, if the conflicts can be fixed up. I'm glad to test out to see if it affects GEOSgcm in anyway. |
Not sure how/why the Base_base stuff got mixed into this branch. Presumably we want all the other stuff. I'll try a hand merge. |
OK- made an attempt. I had to do a new branch "3rd-try" because git would not let me push to origin after I rebased. Going to try to edit this PR, but may have to do a new one. |
Description
Fixing numerical bug in HorizontalFluxRegridder
Related Issue
Motivation and Context
bugfix
How Has This Been Tested?
compiles on Intel
is not exercised by GEOS
Types of changes
Checklist: