Parameter changes: The ACCESS-NRI configuration includes many parameter changes. However, most modifications involve omitting default values, meaning the science should remain unaffected. For detailed comparisons, please refer to the cosima-twg-meeting-minutes. More explanations and issue sources can be found in Section 5 of ACCESS-OM2-1-025-010deg-report.
JRA55 forcing: The COSIMA configuration uses version 1.4, which contains a known issue related to incorrect cyclone representation. But this won’t affect RYF9091 configuration. More details can be found in cosima-twg-meeting-minutes.
Minimum support is available for this outdated configuration from ACCESS-NRI.
The ACCESS-NRI OM3 0.1deg configuration will not be available for quite a while.
Objective:
The goal is to provide a new 0.1deg RYF reference run using ACCESS-NRI executables and configurations initialised from restart files of COSIMA runs. This enables a clear workflow for which COSIMA users can run perturbation simulations using 0.1 deg OM2 RYF.
Would you use this updated workflow in your 0.1deg RYF perturbation experiments?
Yes, I would use the new workflow
No, I wouldn’t
Others
0voters
Since I couldn’t modify the existing poll, I created a new one based on @adele157’s suggestion. This updated poll includes an additional option: “Yes, but only with BGC and WOMBAT upgrades.”
Would you use this updated workflow in your 0.1deg RYF perturbation experiments?
I would fully support this if it includes the updated WOMBAT with fixed iron. We have been wanting to run a BGC RYF control for a while. @pearseb how is that going?
Otherwise I don’t think it’s worth the resources.
A BGC enabled rerun with the improved WOMBAT would be great. There are a few people at IMAS starting to look at impacts of sea ice changes on the Southern Ocean carbon cycle, so a trustworthy control run that we can branch off and perturb would be excellent.
WOMBAT-lite is basically done, the dx/optimization/validation preprint should be up on Biogeoscience very soon.
There are a few options that aren’t included in that base configuration that the community may/may not want want to be included.
Saturation state dependent PIC dissolution
Carbonate chem dependent biotic PIC precipitation
Particle ballasting form CaCO3
Implicit zooplankton functional diversity. (i.e grazing params vary with Phyto concentration (a proxy for size of the organisms therein).
Trade offs being more mechanistic complexity/proccesses involved vs more parameter uncertainty (noting these weren’t included in the core optimisation). Including the more sophisticated PIC is probably a must for any Alkalinity perturbation though.
WOMBAT-mid will probably be ready some time next year and open the door to a lot of additional science with, amongst other thing, multiple P and Z.
So from a BGC perspective, I think it depends a bit on the opportunity cost. And when would be the next time a high res BGC run would be affordable?
If I could only pick one 0.1 degree BGC run in, say, the next 3 years, I would vote to wait and do a big IAF run with WOMBAT-Mid. That would enable a lot of really cool mesoscale bio-physical interaction work that a lot of folks are interested in.
But if 0.1 degree BGC run is affordable every, say 1-2 years, then I think a RYF run with WOMBAT-Lite now could be a good idea. Especially since it is likely perturbation experiments (requiring more runs) could be better suited for -Lite than -Mid.
Personally (to now answer the actual question) I think most of the CDR perturbation experiments I am interested in will be at coarser res (to do a lot) or require WOMBAT-mid (to move beyond a 1P-1Z framework), but I could theoretically envision designing something around the proposed run, resource permitting.
TLDR: I think a high res + new BGC is a big priority, but if this is a twice in a decade opportunity, a larger discussion about waiting for -Mid or whether an IAF would be more valuable could be worthwhile.
@tylerrohr22, @pearseb how many extra tracers do WOMBAT-Lite and WOMBAT-Mid have? Extra tracers are super expensive in MOM5 (ACCESS-OM2), so it’s possible we may be limited to WOMBAT-Lite here? But in MOM6 tracers are much cheaper due to the separate (longer) tracer time step, so perhaps WOMBAT-Mid would be better to wait for ACCESS-OM3-08?
(As an aside: what do we even call a 1/12th deg?!?! Is ACCESS-OM3-12 better than ACCESS-OM3-08?)
Tyler is right. WOMBAT-lite is ready and the iron cycle is much improved. I am just about to submit the paper describing the updates and optimisation to Biogeosciences.
That said, WOMBAT-lite is a moving target and there has been more updates outside what will be presented in the paper. These are:
CaCO3 precipitation, ballasting and dissolution dynamics.
Variations in zooplankton grazing efficiency
an even more updated Fe cycle more in line with Tagliabue et al., 2023 Nature.
Two new tracers: DICr (remineralised DIC) and DICp (preformed DIC)
@adele157 WOMBAT-lite has 15 tracers right now, but this can be reduced down to 12 and the same functionality will be achieved. WOMBAT-mid will have roughly double that.
WOMBAT-lite has 15 tracers right now, but this can be reduced down to 12 and the same functionality will be achieved. WOMBAT-mid will have roughly double that.
@pearseb Have you tried running WOMBAT-Lite in ACCESS-OM2-01 yet? Is it even possible? How expensive is it?
All development has been done in a 1 degree ocean.
The good news is that now WOMBAT-lite (and all WOMBATs) will be functioning in generic tracers, we can apply exactly the same code to different configurations very easily. WOMBAT-lite is, for instance, running well in the ESM1.6 right now. So we should be able to test this soon, but I’ll need someone to help me get it up and running because I do not have the capacity at the moment.
Maybe we should move to a nominal resolution in km?
Like ACCESS-OM3-25K and ACCESS-OM3-8K, and we could then refer to them as 25km and 8km resolution models …
This seems like it could be a high priority for COSIMA to get WOMBAT-Lite running in ACCESS-OM2-01, given all the enthusiasm above for a BGC-RYF. What kind of help do you need @pearseb? Like someone to run the tests for you, because you don’t have time? Or tech support to get it running?
I need someone to help me set up the configuration: inputs files, run directory, etc. I’ll worry about the code.
But, let’s start with 0.25 degrees?
Also, @dkhutch has been playing with the Anderson accelerator to speed up the BGC simulations. Dave, how has this been going? We will need something like this to optimize WOMBAT in high res, because doing my optimization techniques requires hundreds of simulations.
Hi Pearse,
I was meaning to check in with you about testing out the Anderson acceleration on some WOMBAT spinups.
So far I’ve tested age tracer only. It works pretty well so far. The algorithm requires a repeat cycle of simulation (i.e. exactly reproducible circulation run over and over.)
I’ve tested a 1-year cycle, a max 50 iterations of the algorithm, and that gets age to spin up about 10-15x faster than online simulation. Convergence is good - can get the age from 0 to full equilibrium in 250-300 years. But, the 1-year repeat cycle yields age results that are a bit in disagreement with the online age simulation (which I ran for 5000 years).
I tried again with a 10-year repeat cycle of simulation, and fewer iterations (up to 10 only). That is nearly finished. Results look better compared to online age, but convergence is a bit slower… more like a speedup factor of 8x.
Here is a time series of spinning up age on a 10 year cycle. age_series_decade.pdf (15.4 KB)
Here is a comparison with online 5000 year simulation at 2000 m depth. age_2000.pdf (263.1 KB)
The Anderson Acceleration age overshoots the online values a bit, but I think it’s an acceptable solution, and the structure is pretty close to online age.
