Yassir Eddebbar (Aug 26 2020 at 01:43): Yassir Eddebbar (Aug 26 2020 at 01:43):I noticed tracer budgets as an item on the Hi-Res-CESM-analysis diagnostics package list and thought to share initial attempts and issues found while working with the 5 year CORE-forced 0.1º runs.
The main issue so far for demonstrating budget closure is that not all budget terms are available for all tracers. e.g. for O2, all budget terms were available in the Annual Cycle CORE-forced runs except TEND_O2. @Matt Long pointed out that TEND_O2 is now saved as output in the current runs in progress, but HDIFE_O2 and HDIFN_O2 don't seem to be. The latter's contributions are generally small and I concur that they are not a priority given storage constraints, but I wonder if it is worth or feasible saving a short (~month-long) period with all budget terms output to demonstrate budget closure? or perhaps just for one tracer but for the whole run?
Yassir Eddebbar (Aug 26 2020 at 01:46):Here is an O2 budget notebook adapted to outputs from the 0004 run in progress, showing the O2 budget "visually" closes, with a residual that has the expected spatial patterns and magnitude of HDIFE_O2 and HDIFN_O2 (based on the CORE-Forced runs). I have computed the budget terms via simple xarray.roll operations as well as using XGCM with the metrics functionality. Besides some minor differences along the edges, they yield similar results. A review and comments/suggestions on the notebook and/or methods are very welcome.
Keith Lindsay (Aug 26 2020 at 02:01):@Yassir Eddebbar , thanks for chiming in.
My rationale for excluding the HDIFE and HDIFN terms for tracers is that because the mixing coefficients are constant in time, these terms can be computed exactly offline from the tracer values. That's the conclusion I came to from inspecting the code.
Are you saying that when you attempt to close the budget with HDIFE and HDIFN computed using xgcm, that the mismatch is similar to omitting HDIFE and HDIFN. That sounds worrisome. But maybe I'm misunderstanding what you meant.
Yassir Eddebbar (Aug 26 2020 at 17:02):@Keith Lindsay , sorry I should clarify. The HDIFE and HDIFN terms were not included/calculated in XGCM, so the mismatch being similar to omitting them is expected since they were omitted in the budget sum in the notebook, so I don't think this is worrisome.
Yassir Eddebbar (Aug 27 2020 at 15:33):Yassir Eddebbar , thanks for chiming in.
My rationale for excluding theHDIFEandHDIFNterms for tracers is that because the mixing coefficients are constant in time, these terms can be computed exactly offline from the tracer values. That's the conclusion I came to from inspecting the code.
@Keith Lindsay that sounds good. Are the mixing coefficients spatially variable? Happy to jump on a call to discuss more.
Last updated: May 16 2025 at 17:14 UTC